North-Asiatic Languages
with a Special Eye to the Altaic Affiliation Question

> Literature references and annotations by Dick Grune,
Last update: Thu Feb 16 22:16:24 2023.

These references and annotations were originally intended for personal use and are presented here only in the hope that they may be useful to others. There is no claim to completeness or even correctness. Each annotation represents my understanding of the text at the moment I wrote the annotation.
> No guarantees given; comments and content criticism welcome.

Some Notes==== ==== ==== ==== ==== ==== ==== ==== ====

Terminology I:
1. Turkic is composed of Common Turkic (from German: Gemeintürkisch) and Bulghar/Chuvash.
2. Proto-Turkic is composed of proto-Common Turkic and proto-Bulghar/Chuvash.
3. Some people call "proto-Turkic" "pre-Turkic" and "proto-Common Turkic" "proto-Turkic".
4. Old Turkish is the oldest form of written Turkish.

Terminology II: There are a number of words that have representatives in Common Turkish (CT), Chuvash, Mongolian, and sometimes Tungusic, that look as if they might be cognates (note the "might be"), and in which CT s/z corresponds to r in Chuvash, Mongolian (and Tungusic). Examples are:
* CT qısγan- ~ Chuv. xĕrxen- ~ Mong. keri- `to be stingy';
* CT süz- ~ Chuv. sĕr- ~ Mong. sigür- `to fiter';
* CT ekiz ~ Chuv. yəkər ~ Mong. ikire ~ Manchu ikiri `twin'.
These word play a large role in the rhotacism/zetacism debate, but there seems to be no technical term for them. This makes descriptions excessively long, so I call them " words".

Terminology III: The linguistic concept of insubordination plays a role in some of the papers discussed here. It is the use of a subordinate clause as a main clause; this is possible because the main sentence part is not expressed but understood nevertheless. An example in English is: `If you could leave your shoes at the door please.' (omitting the preceding `It would be good'/'I'd like it'.). In English this is considered bad style but in some languages the verb form used in such constructions is fossilized and grammaticalized into a new part of the verbal paradigm.
The term insubordination was introduced by N. Evans, in a chapter in Nikolaeva's 2008 book `Finiteness. Theoretical and empirical foundations'. The term is irksome and confuses people and search engines alike and I cannot think why it has been chosen over the much more appropriate and perfectly unambiguous desubordination. I will stick to the latter term.

I do not know enough of genome techniques to comment on the genome part of a paper. I just accept the conclusions.

End of Notes ==== ==== ==== ==== ==== ==== ==== ==== ====

* Kassian, A.S., Starostin, G., et al., Permutation Test Applied to Lexical Reconstructions Partially Supports the Altaic Linguistic Macrofamily, 2021, pp. 10.
The paper by Turchin, et al. `Analyzing Genetic Connections ...' (2010) is criticized for using doubtful data bases for semantics; unreasonably coarse-grained mapping of the consonants; and doing bilateral comparisons only. The paper by Ceolin `Significance Testing of the Altaic Family' (2019) is criticized for the use of underspecified algorithms; phantom sound correspondences; cherry-picking the word lists; inaccurate use of semantics; and the use of modern languages, some of which, e.g. Khalkha and Manchu, have been in close contact in the last few centuries.
     For Japanese two word lists were used, one with initial d- and one with initial y-, since the experts cannot agree. The proto-Turkic and are interpreted as * and (voiceless) *ɬ, resp., since they appear as r and l in proto-Mongolic, proto-Tungusic, and proto-Korean (either genetically or through language contact), and in Old Chinese, proto-Samoyedic, and proto-Yeniseian (through language contact).
[DG: That is, the authors reject rhotacism/zetacism in proto-Turkic.]
     Basically the algorithm used is the same as that of Turchin et al., with two improvements.
1. Consonants are mapped onto 13 classes, rather than 9; so there are 169 = 13×13 words in the world. This gross approximation is defended on the grounds that this is what comparative linguists do in the initial stages of their work.
2. Matches of semantically more stable notions are counted as more significant.
     The results are that the similarities between the micro-Altaic languages are very to highly significant, unsurprisingly. It is surprising, however, that similarity of Korean with the other languages (including Japanese!) is not significant (reaching a spectacular low of P = 0.577 with Turkic). Also, Japanese is found to be significantly similar to Turkic only (P = 1.8×10^−4 / 5.4×10^−5 for the *d-/*y- versions), not to Mongolic, Tungusic, or Korean (P = 0.094/0.103, 0.004/0.006, and 0.11/0.26, resp.). No significant difference between the *d- and *y- versions of Japanese was found.
     The method does not tell whether the similarities found are genetic or areal, but the authors virtually rule out areal influence between Turkic and Japanese due to distance, and therefore lean towards postulating a genetic relationship.
[DG: Most of the 6000-7000 years of Altaic history the Turks were not in Turkey and the Japanese not in Japan. The distance Yakutsk-Liaodong is about 2300 km, the same as Amsterdam-Athens.)

* Vovin, A., Names of Large Exotic Animals and the Urheimat of Japonic, 2021, pp. 16.
The author identifies the Japanese words kisa `elephant', tora `tiger', and wani `saltwater crocodile' as loans from an Austronesian, more particularly a Palaungic language. This places the homeland of the Japanese in South-East Asia, which fits in with their rice as staple food, houses on stilts, and dwellings that protect better from the heat than the cold.

* Khabtagaeva, B., The Etymological Remarks on Names of Body Parts in Dagur, 2021, pp. 29.
Before embarking on the title subject the author reports on her field work on Mongolian loans in Tungusic languages, in Buryatia in Russia. The conclusion was that all Mongolian loans in the Tungusic languages in that region were from Khamnigan Mongol, and not Buryat or Dagur, as one might expect. That means that the borrowing took place before the Buryat arrived in the region.
     Dagur is a very archaic Mongolic language, and may have split off before Mongolic became Middle Mongolian, i.e. it may be para-Mongolic. The present language, however, is so full of loans from other Mongolian languages and elsewhere, that is not a good representative. To find the "real" Dagur language, and relying on the idea that the words for body parts are quite borrowing-resistant, the author considers 95 body part terms from dictionaries and field work.
     The archaic nature of Dagur is substantiated by 10 sound changes between (Common) Mongolic and Dagur. The first two sound changes express retention of pre-Common Mongolic features (retention of x, which became h in Common Mongolic; absence of regressive assimilation of i to the preceding consonant). The other eight sound changes are specific to Dagur, so they occurred after the split. Examples are rhotacism, the loss of unaccented final syllable, and the palatalization of final consonants. These sound changes relate for example Dagur aurky `lungs' to Mongolic *aguškï.
     All 10 sound changes are supported by numerous examples, but no proto-Mongolic-Dagur reconstructions are given. The paper closes with a list of 95 Dagur words representing body parts, each with its Mongolic equivalent and its provenance. The words were related to: words in Common Mongolic (71); Turkic loans already in Common Mongolic (4); loans from Tungusic (10); and re-loaned Mongolic loan-words in Tungusic (3). Eight words are of unknown origin. These are ńokul `neck'; kuām `neck'; tǝlǝg `body'; mǝk `bosom'; gākā (child.) `hand', arm'; niǝrd `back, spine'; bogōr `buttock'; and xodōs `bladder'.

* Vovin, A., Austronesians in the Northern Waters?, 2021, pp. 29.
Given that the Austronesians, especially those from the Philippines, were proficient seafarers even before the Yayoi crossed over to Japan, it is to be expected that they also visited the coast of Japan. Indeed the author finds traces of North-Philipinic in some `pillow words' in Old Japanese poetry; of Batanic (from an island chain between Taiwan and Luzon) in the words takey `field' and asi `foot' (the latter on shaky grounds); of Amis (Formosan) in isu `whale'; and of Central Philippine in the locative prefix sa-, the purported origin of Old Japanese locative prefix sa-. The author is careful to show that these similarities do not derive from recent Japanese occupation but hail from about 400 AZ.

* Vovin, A., Catching a Black Cat in a Dark Room, 2021, pp. 14.
Review of Robbeets & Savelyev's `The Oxford Guide to the Transeurasian Languages'.
     Very critical review.
Apart from a steadfast belief in the myth of "Transeurasian languages", the biggest criticism concerns the lack of depth, probably resulting from the participation of too many inexperienced contributors: none of the "old" languages, Old Runic Turkic, Old Mongolian, Old Korean, etc. is treated in any depth. This Guide does not guide prospective Altaic experts to the right places.
     There is personal criticism of specific authors too: Robbeets is criticized for introducing the term "Transeurasian"; for the level of her command of the ancient Altaic languages; for the irregularity of her phonetic rules; and for her disregard of actual language data. Savelyev is criticized for hardly mentioning the Bulgharic languages in a chapter named "Chuvash and the Bulgharic Languages"; for engaging in "Bayesian fantasies"; and for putting Khalaj (a -d- language) in the same branch of the Turkic languages as Salar (a -y- language).
     Furthermore the list of basic Altaic vocabulary is taken apart line by line, and in almost all lines errors are identified. Equally serious errors are demonstrated in the table of Korean-Japanese correspondences. Finally Oxford University Press is taken to task for automatically repeating Robbeets and Savelyev's advertisement on the cover; errors in it are pointed out.
     In summary, the book is useless: the good science in it can easily be found elsewhere, and the bad science in it will be harmful to the field.

* Robbeets, M., Savelyev, A. (eds.), The Oxford Guide to the Transeurasian Languages, Oxford University Press, 2020, pp. 976.

Collection of 47 papers on all aspects of the Transeurasian languages, by the experts. As the title implies, the reality of "Transeurasian" seems assumed. No dissident names, here Pakendorf!! ZZ , no Alexander Vovin, no Stephan Georg. The only neutral name is Juha Janhunen.


* Robbeets, M., The Classification of the Transeurasian Languages, 2020, pp. 23.
(Chapter 3 of `The Oxford Guide to the Transeurasian Languages')
     [DG: The classification of languages is expressed by putting their names as labels on the "leaves" of a tree. The leaf labelled with a language sits at the end of a "branch" connecting it to a "node" representing the closest ancestor of that language. This node is again on a branch connecting it to its closest ancestor, etc., until we reach the top of the tree (also called the "root"). Nodes can be labelled with the name of a language, e.g. "Old English" or of a family or phylum, e.g. "West-Germanic", "Indo-European".
     When we consider two languages in the tree and follow their branch sequences until these sequences meet at the same node, we have found the closest common ancestor of the two languages.
     Trees in linguistics often carry time depth as an additional piece of information: the length of the branch is representative of the time depth between two nodes. With all modern languages at the same level (the "now" level) a time scale can be associated with the tree.
     The tree should be binary, but it often is not, implying that a group of speakers split into more than two groups and each went their way. It is not clear to me how likely that is.
     The tree represents the genetic relationship between the languages. But languages borrow from other languages, often from the same tree. This causes cross connections which turn the tree into an acyclic directed graph. This is seldom acknowledged.]
     The author distinguishes three methods of obtaining a classification tree.
1. The classical historical-comparative method; the most closely related languages are considered and sound correspondences between them are established. These are then used to reconstruct their common ancestor; etc. This constructs the tree. Common innovations are placed at the latest possible branch. If written historical material is available this can supply some idea of the time scale involved.
2. Bayesian phylolinguistics; like the above, but given the common innovations, an algorithm is used that considers all possible trees and selects the one that is most likely to produce the observed innovations. The algorithm is based on Bayesian statistics. This takes a measure of human interference out of the classification process.
     3. Lexicostatistics; based on some researcher-supplied measure of the difference between two languages, the tree is constructed that best represents the differences. If we assume that all languages change at the same rate (a shaky assumption) this produces a tree with a time scale.
     For the Transeurasian languages all three procedures give by and large similar results: the top node usually has two children: Turkic-Mongolic-Tungusic, and Korean-Japanese. The structure of the Tk-Mo-Tg node, however, may differ in the place of Tungusic. Also, time depths differ in different analyses, especially in the split between Korean and Japanese (estimates range from 2000 to 6000 years ago). Some researchers arrive at a tree in which the top has three children, while others may not even consider Turkic and Mongolic as a node.
     Using a large database of etymologies and innovations and probably using the program BEAST, the author produces a Bayesian tree of the Transeurasian languages. It indicates that around 5000 BZ Koreo-Japonic split off, which then split in Korean and Japonic around 2000 BZ. Around 3000 BZ Tungusic left the Altaic branch, followed by a split between Mongolic and Turkic at around 1800 BZ.
[DG: This makes Micro-Altaic the same age as Indo-European, but with far less historical material.]

* de Boer, E., The Classification of the Japonic Languages, 2020, pp. 33.
(Chapter 4 of `The Oxford Guide to the Transeurasian Languages')
     Ideally language classification results in a binary tree, each split motivated by a (set of) sign-post lexical and/or morphological and/or phonological properties. This result can be achieved when after the split the language groups each go their ways and they never meet again, which happens ocasionally. But even then "independent parallel development" can cloud the picture; it is important to note that such development is much more likely in morphology and phonology than in the lexicon, the number of items in the latter being much larger. In the Japanese setting near perfect results are obtained on the Ryu-Kyu islands, where split populations stayed split, but on the main islands linguistically split populations remained in contact.
     Proto-Ryūkyūan preserves vowel distinctions that are lost in Central Old Japanese, which shows that proto-Ryūkyūan and Central Old Japanese are sister languages [DG: or COJ is a daughter of proto-Ryūkyūan...].
     After discussing earlier proposals the author sets out to define 13 phonetic innovations (e.g. realization of /i/ as ɨ), 8 tonal innovations (e.g. leftward tone shift), and 11 morphological differences (e.g. negation with -anu/-an versus -anai), and on the basis of these 32 shibboleths constructs a very detailed classification tree of Japonic, containing 48 nodes. The top level has 1. Eastern Old Japanese (with many living descendants); 2. Central Old Japanese; 3. Izumo-Tōhoku; and 4. Kyūshū-Ryūkyū. The tree is completely different from the traditional ones: proto-Ryūkyūan is on level 3 from the top and the Northern-/Southern-Ryūkyūan split is on level 4, the deepest level of the tree.

* Whaley, L.J., The Classification of the Tungusic Languages, 2020, pp. 33.
(Chapter 6 of `The Oxford Guide to the Transeurasian Languages')
     The sound correspondences between the Tungusic languages, as obtained by the comparative method, suggest three different relation trees. Bayesian analysis based on the vocabularies narrows these down to one:

                   +- Even/Evenki/Negidal complex
                   | +- Nanai/Udihe/Oroch complex
                     +- Jurchen, Manchu, Xibe
A more detailed tree is shown in the paper. The split of proto-Tungusic is calculated around a bit before the start of the second millennium.

* Savelyev, A., A Bayesian Approach to the Classification of the Turkic Languages, 2020, pp. 25.
(Chapter 9 of `The Oxford Guide to the Transeurasian Languages')
Due to the high degree of contact between the Turkish tribes, traditional classification yields something that Dolgopolsky once called a genealogical "stump". There are about 4 widely published classifications Turkic; they agree in general outline but differ in detail.
     Bayesian analysis takes a data structure and a set of "a priori" probabilities/beliefs about relationships between components of said data structure, and then produces instances of the data structure with their "a posteriori" probabilities. In our case the data structure is a genealogical tree, and the probabilities/beliefs are linguistic knowledge snippets and their certainties.
     Based on a number of standard data sets and the assumption that evolution speed was the same for all the languages, a DensiTree and the maximum credibility tree for the Turkic family was produced. These are then discussed in the rest of the paper. The DensiTree is vague, probably due to unrecognized inter-Turkic borrowings; most of the nodes in the maximum credibility tree have a credibility of 1, the lowest being the split between Saryg Yugur and the rest of Southern Siberian Turkic, with 0.58.

* Robbeets, M., The Typological Heritage of the Transeurasian Languages, 2020, pp. 54.
(Chapter 10 of `The Oxford Guide to the Transeurasian Languages')

     Answer to Janhunen's "Ural-Altaic: The Polygenetic Origins of Nominal Morphology in the Transeurasian Zone" (2014) in which he claimed that "speaking of “Altaic” instead of “Ural-Altaic” is a misconception, for there are no areal or typological features that are specific only to “Altaic” without Uralic."

     To claim the existence of a "typological heritage" in a "Transeurasian linuistic area" it is necessary to show that the area holds a wignificant number of typological features that do not occur elswhere in signiicant numbers.


* Hübler, N., Typological Profile of the Transeurasian Languages from a Quantitative Perspective, 2020, pp. 48.
(Chapter 11 of `The Oxford Guide to the Transeurasian Languages')


* Pakendorf, B., Aralova, N., Even and the Northern Tungusic Languages, 2020, pp. 42.
(Chapter 18 of `The Oxford Guide to the Transeurasian Languages')
     A parallel grammar of Even (aka Lamut), Evenki, and Negidal, based on the authors' collected oral recordings of Even and Evenki; the Negidal data come from published sources.
     Each of these languages has many dialects: Even 13, Evenki 51, and Negidal 2 (of which one is now extinct). Like all "Altaic" languages they are morphologically relatively simple, but as the Northern Tungusic group they are characterized by large numbers of suffixes, expressing:
15 or so cases for the noun, including two accusatives, one for definite objects and one for indefinite objects;
inclusive and exclusive 1st person plural forms;
10 aspect suffixes, among which an accelerative;
17 mood suffixes, among which remote imperatives for 2sg, 1pl, and 2pl;
12 participle suffixes, among which an impersonal-debitive (probably meaning "someone should ...");
15 converb suffixes, 6 for same subject, 1 for different subjects, 6 for both kinds, and 2 for negative; and
7 valence changing suffixes, including one for adverse-passive.
     There are voiced and voiceless plosives, a fricative β, and difference between l and r. There is both progressive and regressive assimilation. There are 4 vowels, i, e, u, ɵ, which also occur pharyngealized: , , , ; each of these can occur short and long. There is vowel harmony; the details differ per language, and in at least one dialect it is based on pharyngealization. In Lamunkhin Even set 1 consists of e, ɵ, and ie, set 2 consists of a, o, and ia, while i and u are neutral. In many dialects the vowel used with stems with a neutral vowel is lexically determined. E.g. in Lamunkhin Even is- `to tear away' uses suffixes with vowels of set 1, and is- `to reach' uses set 2.
     The authors are sceptical about the genetic relationships of the "Transeurasian languages" and are careful to always write "Altaic" between double quotes.

* Pakendorf, B., , Sakha and Dolgan, the North Siberian Turkic Languages, 2020, pp. 45.
(Chapter 26 of `The Oxford Guide to the Transeurasian Languages')


* Savelyev, A., Chuvash and the Bulgharic languages, 2020, pp. 47.
(Chapter 27 of `The Oxford Guide to the Transeurasian Languages')
     In the 18th century Chuvash was believed to be an Uralic language, but in 1841 it was shown that it is a Turkic language and in 1863 that it, together with the Bulgharic languages, belongs to the Bulgharic branch of Turkic, the other branch being Common Turkic. The branches spilt somewhere around the year 0. The phonetic relationship between the two branches is very complicated and there are hardly any simple regular sound correspondences, only complicated ones.
[DG: up to the point that some (who?) claim the Bulgharic languages form a fourth micro-Altaic branch.]
There is no strong reason to think that the Huns (4th-5th century) spoke a Bulgharic language; the same applies to the Khazars (7th-10th century).
     Chuvash is called čə̂ovaš in Chuvash. It is spoken primarily in the middle Volga River basin, and is written in Cyrilic script. There are two forms, Viryal (North-West) and Anatri (South-East) Chuvash, completely mutually intelligible with a split at around 1700. There are several dialects, e.g. Yenči and Poškărt, the latter highly idiosyncratic.
     Properties of the language.
The plosives are all voiceless, but are semi-voiced between vowels. There is an s (alveodental), an š (postalveolar) and an ś (palatal); and there is an l, an ĺ, and an r, all of restricted occurrence. Consonants in palatalizing contexts are palatalized automatically, if possible. There is a large number of restrictions on co-occurrences of consonants. There are 11 or 12 vowels, in unsystematic relationship; əo is a rounded ə.
[DG: The Wikipedia has 8 or 9 vowels, and əo and ə are not distinguished.]
     Chuvash differs from all other Turkic languages in the order of the possessive and plural suffixes: tus-ə-sem-pe [friend-POSS.3-PL-COM] `with his/her friends', where the plural ending -sem derives from proto-Turkic *sayïn `all'. (Turkish has arkadaş-lar-ı-yla [friend-PL-POSS.3-with]; there is no trace of the Common Turkic plural marker -lAr in Chuvash).
     There are eight cases and four noun classes: ending in r, in other consonants, in ə, and in other vowels; the paper has a 8x4 table of cases X noun classes, with the main surprise that the initial r of an ending turns into t after a noun ending in r: xir-re*xir-te `field-in'. Furthermore there are compound cases, e.g. və̂orma-n-alla [forest-DAT--APPROX] `towards the forest'. There is also quite a number of fossilized cases used with specific nouns only. The combination of cases and possessive markers is far less modular than in Common Turkish and shows much detail. The third person has stand-alone possessive pronouns: un-ə̂n `his/her/its'.
     Like the English `I'/`me', the Chuvash personal pronouns come in two forms: direct and oblique:

		dir.	obl.			dir.	obl.
	1sg	e-pə	man-		1pl	e-pir	pir-
	2sg	e-sə	san-		2pl	e-sir	sir-
	3sg	və̂ol	ə̂on-		3pl	vəosem	vəosen-
Negative pronouns are formed like in English: ni-kam `no-body', ni-mən `no-thing', with negative ni- from classical Persian or Russian, and ta-kam `some-body', te-mən `some-thing'.
     The Chuvash numerals, with the Turkish numerals for comparison:
		Chuvash				    Turkish
	 1 pər		 6 ultə̂			1 bir	 6 altı
	 2 ikə		 7 śičə			2 iki	 7 yedi
	 3 viśə		 8 sakə̂r		3 üç	 8 sekiz
	 4 tə̂ovatə̂	 9 tə̂oxə̂or		4 dört	 9 dokuz
	 5 pilək	10 vunə̂			5 beş	10 on
Most have an alternative "full" form, formed by doubling the middle consonant. Ordinals are formed by suffixing -məš: pillək-məš [five(FULL)-ORD] `fifth'.
     The comparative is formed synthetically: avan-rax [good-COMP] `better'; the superlative is formed analytically: či načar [most bad] `the worst'. Many adjectives can be internsified by partial reduplication: xup-xura [bla-black] `pitch-black'.
     Verb forms are similar in structure to Turkish: kal-ay-ma-r-ə̂m [speak-POT-NEG-PST-1SG] `I was not able to say' (Turkish: konuş-a-ma-d-ım). The verb has 5 moods: indicative, imperative, conditional, concessive, and optative. There are converbs: lar-sa čə̂olxa śïx-at́ [sit-CVB stocking knit-PRS.3SG] `sits and knits stockings'. Modalities are often expressed through participles: un-ə̂n śiy-es kil-et [3SG-GEN eat-FUT.PTCP come-PRS.3SG] `his future eating [be?]comes' → `he wants to eat'. The negative imperative uses a prefix: Muskav-a an=kuś [Moscow-DAT/ACC NEG=move.IMP.2SG] `Do not move to Moscow'. The paper gives many more grammatical details on adjective linking, relative clauses, indirect speech and more.
     The lexicon is for about two thirds identifiably proto-Turkish. There is an early layer of Old Ossetic loans, and many words are from Persian or Arabic origin via Tatar. Loans from Uralic are minor. And for obvious reasons there has been a major influx of Russian terms from administration and modern life.
     [DG: It seems to me that Chuvash is the original Turkish that has missed the regularizing influence of being essentially an army language that made Common Turkish so absurdly regular. This may contradict the Wikipedia, which seems to consider Chuvash an aberrant form of Turkish.]

* Joseph, A., Ko Seongyeon, Whitman, J., A Comparative Approach of the Vowel Systems and Harmonies in the Transeurasian Languages and Beyond, 2020, pp. 23 (486-508).
(Chapter 29 of `The Oxford Guide to the Transeurasian Languages')
     This paper summarizes most of the research into the vowel systems of the North-East Asian languages in a very systematic way. The vowels of the Turkic, Mongolic, Tungusic, and Koreanic languages can roughly be divided in a Set A and a Set B, with the property that vowels of Set A and Set B do not co-occur in words: positively seen this is vowel harmony. In Turkic Set A is characterized by [−back], and Set B by [+back]. In the other phyla Set A is characterized by [−RTR], and Set B by [+RTR] (Retracted Tongue Root).
     Some evidence for the existence of the feature [±RTR]:
1. A vowel with [+RTR] is produced with more tongue-root retraction, as shown in X-ray tracings (already in Russian publications from around 1960).
2. The pharyngeal cavity is larger when saying a vowel with [+RTR].
3. Greater muscular tension associated with [+RTR].
4. In some languages k and g occur only before [−RTR] vowels, and q and ʁ occur only before [+RTR] vowels.
But tests with formant frequencies are inconclusive.
     Some evidence for the claim that Mongolic, Tungusic, and Korean use [RTR] rather than [ATR]:
1. Particles used with neutral stems (stems containing neutral vowels only) almost always have the inactive [−] version of the feature. In Mongolic, Tungusic, and Korean neutral stems use [−RTR] particles [DG: ?? example from Korean, please.]; if these languages used [ATR] these particles wold seem to have [+ATR].
2. In some Mongolic and Tungusic languages the distribution of k and g versus q and ʁ depends on the following vowel. This would be inexplicable if these languages used [ATR] because [ATR] does not affect kand g.
     Turkic has [palatal] vowel harmony and Korean, Mongolic, and Tungusic have [RTR] harmony. Because there is no known phonological way to get from the [palatal] feature to the [RTR] feature (from [RTR] to [palatal] is easy) the conclusion is that Altaic had [RTR] and Turkic lost it in favour of [palatal].
     The authors point out that reinterpreting the [front]/[palatal] dimension of the vowel system as based on the [RTR] feature is not just a renaming exercise, although one might think so because the relations between the vowels and the vowel harmony division are not affected by the reinterpretation. However, the underlying phonetic mechanism is different, so the reinterpretation can explain new phenomena, which may lead to new insights and to changes in the grouping of the languages under their nodes.
     Next Ramstedt's proto-Altaic vowel inventory is reanalyzed in view of our present knowledge of [RTR]. Much attention is paid to the "in-between" vowels like ö, ɯ̬, ü, ÿ, , and ʏ̝, which are most affected by the reinterpretation. Next a similar reanalysis is done on Poppe's proto-Altaic vowels.
     The feature RTR] is amply present in the macro-Altaic languages, with the exception of Turkic and Japanese, but it does not identify these languages as a genetic unit, because the feature is areal, with Eskimo-Aleut, Salishan, and Sahaptian stretching far into North America.
     The proto-daughters of macro-Altaic all have 8 vowels, divided into two groups of four, one with [+labial] and one with [−labial], except proto-Korean, which has six vowels, missing the two [+low][+round] vowels, o and ɔ. At the same time proto-Korean had two vowels that stood out in that they never occurred word-initially, were subject to syncope, and are often called "weak vowels": ɨ and ʌ. There are two ways to explain this imbalance of the Korean vowel system. The first would assume a loss of labial harmony. One would then assume that pre-proto-Korean did have o and ɔ, and that these turned into ɨ and ʌ by losing their [+labial]. This would make the pre-proto-Korean vowel set equal to that of proto-Altaic. The fact that labial harmony decreases as one moves eastwards speaks for this explanation. This is similar to Lee's Korean Vowel Shift, who reconstructs the ancestors of Late Middle Korean ɨ and ʌ as round ([+labial]): *ɔ̈ and *ɔ. The second would be to assume that the Korean vowel system was the original system and that Mongolic and Tungusic developed their labial vowels under the influence of an areal labializing effect advancing from the West.
     Many arguments from Ko `Comparative Consequences of the Tongue ...' (2014) are repeated, often in a more systematic way.

* Robbeets, M., A Comparative Approach of Verbal Morphology in Transeurasian, 2020, pp. 29.
(Chapter 30 of `The Oxford Guide to the Transeurasian Languages')


* Robbeets, M., Basic Vocabulary in the Transeurasian Languages, 2020, pp. 30+voc.
(Chapter 36 of `The Oxford Guide to the Transeurasian Languages')


* Manaster Ramer, A., Duh-Rhotacism, (Turcology 4.01), 2020, pp. 30.
Long, complicated, and sometimes confusing paper, about half of it footnotes, with short in-line literature references (e.g. "Poppe 1974") but no explicit literature references (this one turned out to be `Zur Stellung des Tschuwaschischen'). Also: Spelling checker? What spelling checker?
     Postulates that there are three "Turcologies": Turcology 1, of the pro-Altaicists; Turcology 2, the anti-Altaicists, and Turcology 3, of the agnostics / those lacking courage / etc. The author seems equally disappointed by each: the first and second groups are fanatics, and the third group sit in the side lines and fail to improve the situation. The third group continues to do "astronomy without deciding whether the earth revolves around the sun or vice versa." The author then introduces (writing in capital letters, no doubt to be heard better) Turcology 4.X, remedying the faults of Turcology 1, 2, and 3.
1. One of the features of Turcology [123] was "confusing terminology for substance", and the author investigates all names for proto-Turkic, Chuvash, and Common Turkic that have been (ab)used in the literature. Tables, trees, and quotations remove any shade of doubt that all scholars from about 1920 onwards actually agreed on the substance of this structure and only quibbled about names.
2. The author claims that use of the term "Common Turkic" has led some scholars to consider Chuvash as "aberrant", consciously or unconsciously influencing the debate.
3. Finally, based on a paper cited as "Benzing (1940)" (probably `Tschuwaschische Forschungen I'), the author points out all the misunderstandings in the rhotacism/zetacism debate, emphasizing that r in Chuvash can also come from proto-Turkic d. It is not always clear if the author is expressing his own view or that of his opponent. This makes this section hard to follow.

* Manaster Ramer, A., Arm Chaise Linguistics: A Few Polish Villagers and Guillaume Jacques—and their Implications for Armenian, Turkic, and Uto-Aztecan Linguistics, 2020, pp. 7.
The tone of this paper is aggrieved and sarcastic, sometimes to the point of being incomprehensible, e.g. the half a page footnote #3, and the final paragraph.
     The author points out that whether proto-Turkic had a z or an r in the words is immaterial. If it had an r it still could have inherited it from pre-proto-Turkic (a language unrelated to proto-Whatever) which then turned into z in proto-Common-Turkic, if one chooses to reject Altaic unity. And if it had a z it still could have inherited an r in these words from proto-Altaic, which then turned into z and r in proto-Turkic, if one chooses to accept Altaic unity.
     The arguments in the rest of the paper are more fully expressed in the author's paper `Duh-Rhotacism (Turcology 4.01)' (2020).

* Hudson, M.J., Nakagome, S., Whitman, J.B., The Evolving Japanese: The Dual Structure Hypothesis at 30, 2020, pp. 14.
The scenario developed by Kazurō Hanihara and others in the early 1990s, in which the main population of present-day Japan descends from the Yayoi immigration in the centuries around the year 0 (involving 400.000 to a million people), whereas the Ainu are the remainder of (one of) the peoples of the Jomon culture, has held up well over the subsequent two decennia.
     Some new discoveries:
1. DNA reveals a large Southeast Asian component in the Jōmon people, which shows that they are not direct descendants of the Paleo-Siberian people that must have populated Japan over the Sakhalin land bridge about 20,000 years ago.
2. The Ainu are genetically close specifically to the peoples of northeast Siberia, such as the Itelmen and Chukchi, rather than to those of central Siberia. This genetic link is relatively recent (5th century AZ) and may have resulted from intermarriage between Okhotsk and Ainu people.
3. The genetic relation between Ainu and Ryukyu islanders is much closer than than that of either of them with modern Japanese. This means that the Japanese language spread to the Ryukyus more by language replacement than by population replacement.
     On the linguistic side the authors observe that with the publications of Robbeets (2005) and Francis-Ratte (2016) the evidence that Japanese and Korean are genetically related is gaining acceptance.
     The authors lament that research on the history and languages of Japan by Japanese researchers is still ruled by nationalism and politics (e.g. Tatsuo Kobayashi (2018)).

* Janhunen, J., The Differential Diversification of Mongolic, 2020, pp. 29.
Languages may change along several axes: lexicon, morphology, syntax, numerals, etc. The rate of change in the about 16 Mongolian languages since the 12th century is examined and found to vary widely. For example, of 452 core Old Mongolian lexical items, Khalkha (the official Mongolian language) preserved 95%, against Mangghuer (spoken by approximately 25,000 people in China's northwestern Qinghai Province) only 39%. Morphologically the differences in change rates are not so large. In phonology, however, Khamnigan has undergone about 11 changes since proto-Mongolian, whereas Khalkha has undergone at least 21 changes.
     The author identifies eight factors that may influence the rate of change: geographical position, topographical complexity, economic orientation, population size, age structure, political organization, the existence of written languages, and bilingualism. Although the author examines each of these factors in depth, no relative importance could be assigned.
     [DG: I suppose this effect is not specific to Mongolian. This puts glottochronology on very shaky footing!]

* Vovin, A., Old Korean and Proto-Korean `r' and `l' Revisited, 2020, pp. 14.
The author observes that -ri- in Old Korean sometimes shows up in Middle Korean as -ri- and sometimes as -i-. There are 10 words in which the `r' is lost (9 nouns and 1 morpheme), and 4 words in which the `r' stayed (1 verb, 1 noun, 1 pronoun, and 1 morpheme). (The difference in size between the two sets is not significant, and caused by the paucity of Old Korean material: there are many words with -ri- in Middle Korean, they just are not represented in the Old Korean material.) The most important item is the morpheme in the second set, -ri = irrealis participle -l + nominalized i. So apparently it is the `l' that stays and the `r' that gets lost. This may be an areal phenomenon: Tungusic Neghidal, presently spoken in the estuary of Amur river, has ry|ε|l. This means that Korean wuri 우리 was originally pronounced /uli/, and invalidated the purported cognate pair proto-Turkic bir² ~ Old Korean uli `we'.

* Yurayong, C., Szeto, P.Y., (De-)Altaicisation as Convergence and Divergence between Japonic and Koreanic Languages, 2020, pp. 16.
On the (implicit) assumption that the similarities between Altaic, Japonic and Koreanic are due to areal rather than genetic influences, and on the (equally implicit) assumption that structural features are exchanged when languages are in contact, a list of 40 structural features relevant to the North-Asiatic languages is drawn up. The languages include Japonic, Koreanic, Northern Sinitic, Southern Sinitic, Core Altaic (Mongolic, Tungusic, Turkic), Ainuic, North-east Asian (Yukaghir, the Chukotko-Kamchatkan languages, Naukan Yupik (Eskimo-Aleut), Nivkh), Russian, and Atayal (Formosan). A relation matrix of languages X features was set up and fed to the program NeighborNet, yielding a plot of the typological distances between the languages.
     The program identified, clockwise, the following neighbours: Core Altaic, Russian, North-east Asian, Koreanic, Japonic, Ainuic, Northern Sinitic, Souther Sinitic, and Atayal.
     This makes the North-east languages (in particular Nivkh) and Japanese the feature-wise nearest neighbours of Korean, and Korean, Ainu, and Northern Sinitic the feature-wise nearest neighbours of Japanese, with the closest proximity between Korean and Japanese. But this proxinity is less than that between the separate languages in Core Altaic. This corresponds well with what is known of the history of these kangauges, and where they were physically neighbours.
     With respect to Korean and Japanese, the authors conclude from the typological distance plot that there was a converging period of perhaps a 1000 years, Altaicizing Japonic and Ainuic [DG: How?] and de-Altaicizing Koreanic, resulting in a new Japonic-Koreanic type of grammatical system. Next a diverging period of perhaps a 1000 years created the present-day Korean and Japanese languages.

* Wang Chuan-Chao, Robbeets, M., The Homeland of Proto-Tungusic Inferred from Contemporary Words and Ancient Genomes, 2020, pp. 12.
Three methods for determining the home land are employed: diversity hotspot, matching native words for agricultural and wild plants and trees to the locations where they are found, and matching genome information from finds in caves from perhaps 6 millennia ago to that of present-day Tungusic-speaking population. The areas produced by these methods overlap around Lake Khanka, (today) a 30x60 km lake on the border between China and Russia, 200 km from the sea of Japan and 250 km north of the North-Korean border.

* Robbeets, M., The Transeurasian Homeland: Where,What and When?, 2020, pp. 38.
(Chapter 44 of `The Oxford Guide to the Transeurasian Languages')

     Based on data from the languages, genetics and archeology, the author concludes that
1. the original speakers of Proto-Transeurasian lived along the West Liao River Region around 6200-4500 BZ;
2. around 4700 BZ the Proto-Japono-Koreans split off and moved south to the Liaodong Peninsula;
3. around 3300 BZ another group split off, moved north-east to Amur-Sungari-Ussuri Basin and became the speakers of Proto-Tungusic;
4. around the same time the remaining group moved north [DG: pushed by the encroaching desert?] where they met the southern tip of the Greater Khingan Mountain Range; those that settled the eastern flanks became the Mongols, those on the western flanks became the Turks.
5. again around 3300 BZ the Proto-Japono-Koreans split and the Koreans moved to Korea;
6. the Japanese also moved to Korea and from there to Japan (200 BZ - 300 AZ).
     [DG: This scenario has the Koreans and the Japanese living side by side on the Liaodong Peninsula and in Korea for about 3000 years. This is not credible. Korean and Japanese are much too different, certainly when we have to believe that they were the same to start with. This cannot be corrected by minor tweaking.]

* Jeong Choongwon, Wang Chuan-Chao, Ning Chao, Transeurasian Unity from a Population-Genetic Perspective, 2020, pp. 10.
(Chapter 45 of `The Oxford Guide to the Transeurasian Languages')


* Li, Tao, Transeurasian Unity from an Archaeological Perspective, 2020, pp. 14.
(Chapter 46 of `The Oxford Guide to the Transeurasian Languages')


* Hudson, M.J., Language Dispersals and the `Secondary Peoples’ Revolution': A Historical Anthropology of the Transeurasian Unity, 2020, pp. 24.
(Chapter 47 of `The Oxford Guide to the Transeurasian Languages')


* Hölzl, A., An Unknown Manchu Dialect?, 2020, pp. 5.
A 69 words long wedding song written in Chinese characters is analyzed and found to be in a language close to Manchu.

* Georg, S., A Modest Personal Remark on Rhotacism: New Abstract, 2020, pp. 1.
Another low point in the Altaic debate. The author feels compelled to issue a formal statement to the effect that he has never said that the sound change rz was impossible [DG: nor that there were angels dancing on the tip of that needle, I suppose].

* Savelyev, A., Jeong Choongwon, Early Nomads of the Eastern Steppe and Their Tentative Connections in the West, 2020, pp. 17.
There is no unambiguous archaeological, historical, linguistic or genetic evidence that the European Huns are a continuation of the Xiong-nu. neither is there any evidence to support a specific connection between the Avars and the Asian Rourans.

* Gakuhari, T., Nakagome S., et al., Ancient Jomon Genome Sequence Analysis Sheds Light on Migration Patterns of Early East Asian Populations, 2020, pp. 10.
Genome material from a 2500-year-old individual from Jomon Japan shows signs of stemming originally from South India, then South-East Asia (with affinity to Taiwan), then North-Eat Asia, and finally Japan. The split with Native Americans was at 23 kya (kilo-years before present).

* Robbeets, M., The Oxford Guide to the Transeurasian Languages: Introduction, 2019, pp. 6.
(Draft of the Introduction of `The Oxford Guide to the Transeurasian Languages')


* Jeong Choongwon, Balanovsky, O., et al., The Genetic History of Admixture across Inner Eurasia, 2019, pp. 13.
[Summarizing material on a subject I know next to nothing about.]
     Genome data of 763 people from 91 ethnicities from all over Inner Eurasia was compared and correlated with data from two people from the Eneolithic Botai culture in Kazakhstan, from about 3,400 BZ (the Ancient North Eurasians (ANE) component).
     The analysis shows that three groups can be distinguished in Inner Eurasia, along geographical lines: the northern-most ‘forest-tundra’ group, with people of ANE ancestry; the intermediate ‘steppe-forest’ group with people of Late Bronze Age steppe descent; and the ‘southern steppe’ group, with people with strong West/South Asian influence.
     Eastern Eurasian ancestry is especially pronounced in the Nganasans (Samoyedic); they may be the isolated descendants of a prehistoric Siberian population with high ANE affinity.
     The information is summarized in a large map of Eurasia and two detailed maps, of the Caucasus and of the Altai-Sayan region. Selected genetic information from all participants is given in colourful bar charts, grouped by ethnicity; even to the layperson one group stands out as being completely different: the Amis (Taiwanese aboriginals).
     Unfortunately not all abbreviations are explained. What is `Okn'? Okinawan?

* Robbeets, M., Janhunen, J., Savelyev, A., Korovina, E., The Homelands of the Individual Transeurasian Proto-Languages, 2019, pp. 55.
(Draft of chapter 43 of `The Oxford Guide to the Transeurasian Languages')
     Several dozen Chinese loans in Old Turkish, mainly war terms, plus other considerations place the Turkic homeland between the Sayan-Altai Mountain region in the west to present-day Inner Mongolia and Shanxi in the east, at about 3000 BZ.
     The modern Mongolian languages all derive from the language of the Mongol Empire (1206–1368), and its homeland was the region between the Onon and Argun rivers. But there are other (Macro-)Mongolic languages: extinct Khitanic, and perhaps others. Agricultural terms suggest southern Manchuria as its homeland, and if it can be identified with the Hongshan culture, the time would be about 4500–2900 BZ.
     Using the "hotspot" criterion of language diversity, the authors locate the homeland of the Macro-Tungusic languages around Lake Khanka (ca. 3200–1300 BZ).
     The ultimate homeland of Japano-Koreanic may have been the Bohai coast and the Liaodong peninsula, which the Koreans left somewhere around 3500 BZ for the Korean peninsula.
     The Japanese remained behind on the Liaodong peninsula, where they may have been in contact with Austronesian-speaking people migrating from Taiwan. They left for the south-western part of the Korean peninsula somewhere before 1500 BZ, to arrive there before 900 BZ, when they started to cross to Japan.
     All arguments are supported by word comparisons from agriculture, mainly concerning millet. There are maps for each of the homelands.
     [DG: Footnote 3 tells us that co-author Janhunen does not accept Japano-Koreanic. Also, from page 35, "Bayesian inference estimates that Japonic and Koreanic separated around 1847 BZ, while lexicostatistic methods yield the 4th millennium BZ". This is a discrepancy of no less than roughly 2000 years, casting more doubt on the existence of Japano-Koreanic.]

* Ceolin, A., Significance Testing of the Altaic Family, 2019, pp. 32.
The author argues that one should concentrate on lexemes rather than on morphemes for significance testing of language relationships.
     For the statistical tests words from modern languages are used; although this is an unconventional choice, it avoids using perhaps controversial etymologies. Words are tested on their first consonant only: if there is a relation between two Altaic languages in any way, it should show up in the first consonants, because the Altaic languages are low on prefixes.
     Several Swadesh lists were combined, and if a word had a native representation in Turkish and in Mongolian and in Manchu, it was added as an item to the test list. Onomatopoeia, nursery words, etc., were removed. This left a list of 168 items.
     Three published statistical algorithms were tested for usability for gauging languages relationships, by running them on word lists of sets of known related languages (English ~ Italian ~ Hindi) and known unrelated languages (Turkish ~ English, Manchu ~ Hindi, etc.). The p-values (likelihood that the lists were random) for the related set was in the order of 0.01 to 0.001, depending on the algorithm, and about 0.5 for the unrelated set. This shows that the algorithms work properly on word lists.
     The sensitivity to loan words is determined by artificially mixing in loan words.
     Although the results of the three algorithms do not exactly coincide (reasons are explained in the paper), the over-all result is that a non-significant similarity iis detected between Turkish and Mongolian, and between Mongolian and Manchu, but not between Turkish and Manchu.
     [DG: Given the areal configuration, it is likely that any similarity detected here is from very old loans, which masquerade as native words.]

* Unger, J.M., The Origins of the Korean Language: A Linguist’s Perspective, 2019, pp. 17.
Starts with a 5-pages crash course in the comparative method, from phonemes to language trees, followed by a warning against politically motivated theories about language origins.
     Modern research (in particular Francis-Ratte, 2016) has resulted in hundreds of credible correspondences, which cannot be attributed to borrowings, because "only three or four dozen words can be securely identified as borrowings from Old Korean into Old Japanese in the 7th and 8th centuries". This proves that Korean and Japanese are related.
[DG: This does not prove anything. There can be older, more disguised, borrowings or even very old borrowings.]
Some examples of lexical correspondences are shown, followed by phonetic rules. Examples of obvious loans are given, showing that these do not obey the phonetic rules (all from Ratte). (The paper does not mention Robbeets.)
     Based on this, the author proposes the following scenario. Roughly 5000 BZ Macro-Tungusic was spoken on the west coast of the Bay of Bohay. Around that time Proto-Tungusic split off and the speakers moved to near lake Khanka. Those that remained behind spoke Proto-Korean-Japanese and split in a northern group, speaking pre-Korean and a southern group speaking pre-Japanese. The northern pre-Korean group moved east and then south to the east coast of Korea, where they arrived around 1500 BZ, creating the Mumun culture. The pre-Japanese group followed the southern coast of the Bay of Bohay into the Shandong peninsula, from where they crossed over to the Liaodong peninsula (50 km) and from there to east Korea, and some hundreds of years later to Japan.
     The author thinks it likely that the Koguryo kingdom spoke Old Korean. The same holds for the rulers of Paekche, but the population may have spoken some some para-Japanese language.

* Hölzl, A., Hölzl, Y., The Endangered Languages of the Manchus, 2019, pp. 54 (slides).
After summarizing the state of the Manchu language (a few speakers left, and perhaps 30,000 speakers of Siba in Xinjiang), the authors report on a field trip to some villages in Guizhou, where people of Manchu descent live.
     The Manchu of Guizhou are aware of their descent, but speak Chinese. Nobody can speak Manchu, but some try to learn it from DVDs. Two families have genealogy books, tracing them to Manchu arriving in Guizhou in the 17th century. Tombstones were visited and investigated for Manchu inscriptions, but little material was found.
     There is also a wedding song from this region (see Hölzl (2020)); it is given with translation. Three word lists purported to represent a Manchu language, with Chinese translation, are available from the middle of the 20th century. These lists were combined and analyzed by the authors, but many words could well be from a Sino-Tibetan language. The other words require further study. The full lists are included, but largely without English translation.

* Elmer, P., On the Origins of the Japanese Language, 2019, pp. 28.
Based on the author's thesis "Origins of the Japanese Languages--A Multidisciplinary Approach".
The author takes stock of the different scholarly opinions on the origins of the Japanese language.
     The southern connection assumes that some Austronesian language was present in Japan before the Yayoi arrived (300 BZ), and that this Austronesian language remained as a substratum in Japanese. Kawamoto (1977) produces several hundreds of etymologies between Japanese and proto-Austronesian. Benedict (1990) declared these etymologies "look-alikes" and placed Japonic as a branch of his Austro-Tai macro-phylum. Blust (2013) declared virtually every etymology between Japanese and Austronesian as problematic. Archeology shows that Austronesians reached the southern Ryukyu islands only.
     The northern connection holds that Japanese is part of the Altaic macro-phylum. This hypothesis is supported by many scholars and reached it peak in the "Etymological Dictionary of the Altaic Languages" by Starostin el al. (2003), with about 1500 etymologies, and in the work of Martine Robbeets (2005), with about 350 vetted etymologies.
     Again these etymologies were heavily criticized, and the critics claim that the similarities between Japanese and Korean are better interpreted as due to convergence (borrowing) than to divergence (from a common source).
     Numerous attempts have been made to combine the northern and southern hypotheses, and make Japanese an Altaic language with an Austronesian substratum.
     The first moment Japanese shows up in written history is in some tens of names and titles of the Wa people in old Chinese documents. The Old Chinese text is vague about date and place, but suggests that the Wa lived to the south of the Kaema plateau in North Korea, about 350 BZ; another Old Chinese text places them next to the Yuè in southeast China.
     The latter would tally well with the idea that the Yayoi people brought wet rice agriculture from southeast China to Japan, which allowed them to expand rapidly there. Various scenarios are described, and linked to myths, which sometimes combine elements from northern (Altaic) and southern (Austronesian) influences.
     DNA has shown that the Hayato people were not indigenous (i.e. part of Jomon) but early Yayoi, so they may have spoken a form of Japonic. Suggestions that they were the ancestors of the Ryukyu people fail on the late arrival of the Ryukyu languages on the islands.
     All in all the author's examination of archeology, agriculture, myths, and DNA research in the region is compatible with the farming/language dispersal hypothesis scenario: The ancestors of the Yayoi people learned wet rice cultivation in an area south of the Yangtze River from an Austronesian-speaking people. The travelled north to the north of the Korean peninsula, where they came into contact with Koreans who spoke an Altaic-type language and who practiced millet cultivation; place names demonstrate their presence in the Korean peninsula. From the beginning of the first millennium BZ, the this people started to move to Japan, where they overran the Jomon. The first to cross were the Hayato, the bulk of the immigrants were the Yayoi. This scenario makes it likely that Japanese and Korean are "two different languages".

* Vovin, A., A Sketch of the Earliest Mongolic Language: The Brāhmī Bugut and Khüis Tolgoi Inscriptions, 2019, pp. 36.
The language of the KhT and Bugut steles is beyond any doubt Mongolic, and from the timing it is reasonable to formulate the hypothesis that it was the language of the Khaganate that immediately preceded the first Turkic khanate. It is known the language of that realm was named Ruan-ruan (or Rouran, etc.). The author acknowledges in a footnote that this leaves nameless the non-Altaic words attributed by him to Ruan-ruan in previous publications.
     It is very satisfying to see that the initial p reconstructed for Middle Mongolic initial h indeed shows up in Ruan-ruan: RR paran ~ MM haran `person'. There is no palatalization of ti and di yet: RR bitig ~ MM bičig `inscription' (from Mon. бичих `to write'). There are very few initial nasals; that may be coincidence or Turkish influence. Although s and sh are distinguished in the script, the sh symbol occurs only in front of i and s is used elsewhere. There is vowel harmony, probably palatal. Plurals can be made with -nAr/-ńAr or -d, as in Khitan and Mongolian, but the distribution is different. The word order is predominantly SOV but OSV also occurs. Many other details of the language are derived from the text, for example the several forms and uses of the genitive (possession, connecting an adjective to a noun, the actor of a passive participle).
     A vocabulary of 113 items, a first translation of the Bugut inscription, an improved translation of the Khüis Tolgoi inscription (Stone One), and the translation of two very short inscriptions from Keregentas close the paper.
     This makes Mongolic the oldest recorded Altaic-type language, taking the place of Old Turkic.

* Lévêque, D., Description of Verbal Morphology of Asama: A Realizational and Implemented Approach, 2019, pp. 4.
A too short description of a very interesting non-Altaic looking Japonic language, in which morphology is often expressed by consonant modification and/of vowel length rather than by suffixes, and in which V:CV and VCV: often occur in free variation.

* Satoshi, M., Nivkh Loanwords in Japanese and Korean, 2019, pp. 7.
Lists 43 etyma, with minimal comment; 33 are to Old Japanese (possibly through a form of Korean), and 10 are to Middle Korean. Examples of proposed loans to Middle Korean are:
proto-Nivkh *baʀ `stone' ~ Middle Korean pahwoy ~ Modern Korean pawi „boulder‟;
proto-Nivkh *liɣr, root *liɣ `wolf' ~ Middle Korean ilhi (with metathesis of i and l to avoid word-initial l) ~ Modern Korean ili `wolf';
Nivkh pasq `one of a pair` ~ Middle Korean pcak ~ Modern Korean ccak `(one of a) pair'.

* Hubler, N., The Historical Signal in Transeurasian Structural Features, 2018, pp. 18.
[DG: These are slides without further explanation, so I may have misunderstood.] As an experiment, the author builds a Bayesian tree based on structural features of the Transeurasian languages, and compares it to Robbeets' tree, which contains much more info. Some conclusions: Yakut, in intensive contact with Even and Evenki, does not cluster with these; Kalmyk, geographically distant from all the other Mongol areas, still clusters with Mongol; Korean and Japanese cluster more closely than in Robbeets' tree.
     [DG: The author's name is Nataliia Neshcheret in the slides, rather than Nataliia Hübler.]

* Vajda, E., Transeurasian as a Continuum of Diffusion, 2018, pp. 14.
(Draft of chapter 41 of `The Oxford Guide to the Transeurasian Languages')


* Maue, D., Ölmez, M., The Khüis Tolgoi Inscription, 2018, pp. 17.
[DG: Confusion reigns as the text of the Khüis Tolgoi inscription starts on the second stone (KhT II), which is badly readable, and ends on stone one (KhT I), which is better. Vovin does not agree as his "Interpretation" 2017 concerns stone 2, which starts in the middle of the text. Please, gentlemen.]
     Updated version of Maue (2017), with additions. The steles were discovered in the Tuul river basin, about 200 km west of Ulaanbaatar. The Niri Kagan in the inscription is the Niri of the Chinese sources, who reigned from 595 AZ. The consonant inventory of KhT is compared to that of pre-proto-Mongolic: p has not yet changed to x; t is still t before i; and the status of ñ and h is unclear. A longer list of morphemes is composed. And a more coherent translation is produced.
     The paper does not mention Ruan-ruan.

* Robbeets, M., Bouckaert, M., Bayesian Phylolinguistics Reveals the Internal Structure of the Transeurasian Family, 2018, pp. 18.
Earlier version of Robbeets, "The Classification of the Transeurasian Languages" (2020).

* Vovin, A., Fabrication of Turkic böz ‘fabric’ in Japan and Korea, 2018, pp. 22.
For the rhotacism/zetacism debate see Tekin (1979).
     Highly sarcastic critique of R.A. Miller's `Turkic böz ‘fabric’ in Korea and Japan' (2005) [DG: which I have not been able to access], in which Miller separates Turkic böz `fabric' from Arabic bazz and Classical Greek βύσσος, of roughly the same meaning, creating an all-Altaic word bör with reflections even into Korean and Japanese.
     If the word böz is really of Altaic origin, the r in Chuvash pir (an word) is original and does not derive from proto-Common Turkic s/z. On the contrary, the Common Turkic s/z then derives from a proto-Turkic , whose reflection can be found back in Mongolian and Tungusic. This strengthens the Altaic Hypothesis.
     But, as Georg (2003, `Japanese ...') has pointed out, Turkic böz looks suspiciously like a loan from Arabic bazz or Classical Greek βύσσος, of roughly the same meaning. If this word is indeed a loan, it is not proto-Turkic, and Chuvash must either have borrowed it from Old Turkish or directly from the source. In either case a sound shift z/sr must have taken place in Chuvash to produce the Chuvash form with r. This means that all words could be originally proto-Turkic, and have obtained an r in Chuvash via the same sound shift. Then these words with proto-Turkic z cannot be related to similar words with r in Mongolian and Tungusic. This makes them irrelevant to the Altaic hypothesis.
     The present paper makes it appear as if Miller declares the word to be of Altaic origin and to be a Wanderwort, which is absurd. Anyway, Miller's arguments are refuted by the author in detail.
     The title makes it clear that the author finds Miller's paper pure fabrication.

* Hölzl, A., The Tungusic Language Family through the Ages: Interdisciplinary Perspectives, (slides), 2018, pp. 36.
There are about 20 Tungusic languages, in total spoken by under a 100,000 people over a stretch of about 4000 km in Siberia from the Yenisei river to Kamtchatka. They were once very important in China as the Manchu dynasty. Most are spoken in Manchuria, except, remarkably, Manchu. The time depth of the present language is about 2000 years. The family consists of four subgroups, represented by Manchu/Jurchen, Nanai, Udihe, and Ewen. There is little agreement over the tree structure, but Jurchen/Manchu may have been the first to split off. There has recently been a surge in Tingusology studies, and last few slides list 115 literature references, written in 16 different languages.
     Remains of two women of 7700 years ago have been found in Devil's Gate Cave near lake Khanka that are genetically directly related to the present Tungusic-speaking population. The author warns against hasty conclusions.

* Hölzl, A., Udi, Udihe, and the language(s) of the Kyakala, 2018, pp. 36.
There are two languages Kyakala, one a dialect of Udihe, described in 1908 and (relatively) well-known, and a second one, described in the 1980s by a native speaker in Chinese characters, and which leans towards Jurchen/Manchu, with several layers of borrowings. Complete word lists of both languages are presented, and compared to Udihe, Jurchen, and a few other Tungusic languages.
     Udi, a Caucasian language, has nothing to do with this paper. It is the favourite language of Prof. W. Schulze, to whom this paper is dedicated.

* Vovin, A., Origins of the Japanese Language, 2017, pp. 34.
After demonstrating the relatedness of Tokyo Japanese and Shuri Ryukyuan the author uses six words from these languages to introduce the notion of a `productive-predictive correspondence': a correspondence that allow one, given a word from language A, to predict the form of that word in language B, and possibly vice versa. Such correspondences can be found for the pair `Tokyo Japanese and Shuri Ryukyuan'. The author considers two languages proven related when there are paradigmatic correspondences and/or productive-predictive correspondences, and preferably both.
     To consolidate the notion of `Japonic', the author gives short lists of words that make the relation of pseudo(?)-Koguryo, Silla-Japonic(?), Paekche-Japonic, Karak (the language of Kaya?) and Jejudo(sic!) to Old Japanese plausible. [DG: the analysis of Tanmura, the old name of Jejudo as tan-mura - `folk village' seems far-fetched].
     The only serious candidate for being related to Japonic is Korean, and the only serious attempt to prove the relationship is by Whitman (2012). Whitman gives 9 corresponding morphological elements, and the author explains in the next pages why each of them is unacceptable: comparison of one letter, use of rare or non-existent words, semantic mismatch, the morpheme is specific to adjectives in one language and to verbs in the other, etc. Several of Whitman's lexical correspondences are rejected because they are not predictive. Irregularities in Whitman's rules for *c and *-k are pointed out. Finally a list of 7 basic words is presented, with their translation in Russian, German, French, West Old Japanese, and Middle Korean. It is immediately obvious from this list that Russian, French and German are related and Old Japanese and Korean are not, or rather that if they are related, their time depth is far greater than that of Proto-Indo-European.
     Attempts to link Japanese to Altaic (Starostin, Robbeets) are booed off. The final conclusion is that "Japanese is a member of a Japonic family of languages. It has no demonstrable relationship with any other language or language family on the planet."

* Vovin, A., And So Flows History, 2017, pp. 13.
The author makes two points: 1. language loss is inevitable and a fact of life; 2. languages revitalization projects are doomed. The corollary is that language recording and language description should get priority over language revitalization.
     Three examples are given for point 1:
1. Inner Asia, where we lost Tocharian, Ruan-ruan, Sogdian, some Yeniseian languages, etc.;
2. the Japanese islands, where we lost Hayato, Kumaso, various dialects of Japanese and Ainu, and some Ryukyu languages; and
3. Korea, where we lost Koguryoan, dynastic Paekche, common Paekche, Jurchen, etc.
     Also three examples are given for point 2:
1. The Ainu project revitalization failed under the eyes of the author because the revitalized Ainu was not used for daily communication even between people who spoke it fluently, and because a formal text could not reasonably be translated into it.
2. In Okinawa, the local language was both suppressed by the government and discouraged by parents, who wanted good Japanese-based jobs for their children.
3. For Hawai'ian, there are 300 monolingual speakers; all the rest at least know English. Hawai'ian is state-supported and all official documents are bilingual, but everybody reads only the English version. The only success story in language revitalization is Hebrew.
     In many countries local dialects survive happily side by side with the official language [DG: in Germany in school and at the work place High German is spoken, but at home the local dialect is used], but there is only one dialect in Japan in that position: Western Toyama.

* Miyake, M., Jurchen Language, in Encyclopedia of Chinese Language and Linguistics, 2017, pp. 5.
Jurchen, closely related to Manchu, is known from inscriptions, word lists, and a few texts in two scripts derived from Chinese in the beginning of the 12th century. The scripts are not fully understood and may hold the key to further decipherment of the Khitan language. The texts are from several dialects over a period of 3 to 4 centuries.
     Jurchen is conservative with respect to Manchu in some aspects (e.g. dentals are not palatalized before i) and innovative in others (e.g. Manchu u and ū are merged in Jurchen). Some of the numerals differ from those in Manchu; they may be loans from Khitan. In morphology Jurchen is richer than Manchu: it has more cases and more vowel harmony, resulting in more particles. Syntactically few differences are known.

* Robbeets, M., Proto-Trans-Eurasian: Where and When?, 2017, pp. 27.
Both sides in the Altaic debate agree that the Altaic/Trans-Eurasian [DG: spelled with a hyphen in this paper] languages hail from a relatively small area consisting of Inner Mongolia, southern Manchuria, and Northern Korea, an area of perhaps 500 x 300 km. Archeology shows that from the 7th millennium BZ onwards millet was grown in this area. This is remarkable because Altaic peoples are usually associated with herding cattle on steppes. Following the trail of the East Asian staples can shed some light on this.
     The proto-Trans-Eurasian word *pisə is reconstructed for `(barnyard) millet'. It has cognates in Tungusic as *pise `seed'/`millet', in Korean as *pisi `seed', and in Japanese as *piye.
     Barley reached the region about 2000 BZ. The Trans-Eurasian word for `barley' was borrowed from a branch of Iranian (compare Pashto orbǝša `barley') into proto-Turkic as *arpa/*arba, next into proto-Mongolic as *arbai, from there to proto-Tungusic as *arfa `barley', `oats', and finally into proto-Japanese as *apa `millet'. [DG: Modern Hungarian has árpa.]
     Several Indo-European languages hint at the existence of a PIE root *mṛk `wheat', which by way of Tocharian might be the source of Old Chinese *mə.rˤək ‘a kind of wheat’. Derivatives of this word found their way into proto-Tungusic, proto-Koreanic (Middle Korean 밇), and proto-Japonic *mugi `wheat'.
     The Altaic languages do not have a common word for `rice'. Proto-Japanese, Old Chinese, and the Austronesian languages used nouns derived from verbs meaning `to eat'. A root for `processed rice' is found in Old Japanese as *kəmai, in Austronesian as *Semay, and in Old Chinese as *C.aj `to destroy`/`rice gruel'. The second meaning of the Old Chinese word may have been borrowed by the two other language groups.
     The first evidence for horse-riding in Northeast Asia goes back to the first millennium BZ. The word for `horse', some cognate of *mark, can be found in Indo-European, Mongolic, Tungusic, Korean, Japanese [DG: although proto-Japanese *uma/*muma is a stretch], Old Chinese (as **mˤraʔ), and Nivkh. This wide distribution points a very speedy spread.
     Many details of the above borrowing processes are supplied in the paper.
     DNA places the split of the Yayoi culture people from other East Asian peoples at 2800 BZ, and their first contact with the Jomon people at 2500 BZ.
     It can be tentatively concluded that the speakers of proto-Trans-Eurasian were part of the earliest Neolithic cultures in Northeast Asia, living in eastern Inner Mongolia and southern Manchuria in the 7th and 6th millennium BZ, where they cultivated millet. Around the middle of the 5th millennium BZ they split in a riverine Altaic group and an oceanic Japano-Koreanic group. The latter group split when the Koreans brought millet cultivation to Korea around the middle of the 4th millennium BZ.
     Increasing aridity and a colder climate in the region, combined with pressure from the Sinitic peoples in the south, pushed the Turks west and the Tungus north-east near the end of the 2nd millennium BZ. Similar pressures forced the Japanese farmers in the Liaodong area to migrate to the Korean peninsula around 1300 BZ, whence they moved to the Japanese islands in the 1st millennium BZ.
     Turkic split off in the 4th millennium BZ, and its move westwards was sped up by horse riding and pastoralism, bringing its speakers to present-day Turkey in the 11th century AZ. Pastoralism developed in East Asia not before the 1st millennium BZ.

* Vovin, A., Old Korean and Proto-Korean `r' and `l' Revisited, 2017, pp. 29.
[DG: These are slides, so the narrative is missing.]
     Modern Korean has two endings -ㄹ (both pronounced `l'), one on nouns for the accusative, and one on verbs for the future/irrealis participle (they differ in details, however). In the pre-Hangul semi-phonetic Korean script Hyangchal they are represented by two different Chinese characters: 乙 for the accusative and 尸 for the participle. This proves that they were pronounced differently in the 10th to 12th century. Given the ambiguity of `l'/`r' in East Asia, the obvious suggestion is that they were pronounced `l' and `r'. Yi Ki-mun assumed the accusative 乙 to have had the sound `l', and the participial 尸 the sound `r', based on comparisons with "Altaic" languages. But because the Altaic hypothesis has not been proven, these comparisons are not valid.
     On the basis of the Middle Korean accent and examples from the Hyangka the author argues for participle 尸 pronounced as `l' and accusative 乙 as `r'. Confirmation is found in obvious loans from Middle Korean in Middle Mongolian and Jurchen/Manchu, both languages that differentiate between `l' and `r'.
     The reasons for the loss of distinction between `l' and `r' in Korean are substratal influences of the remaining Japonic languages in south-west Korea, combined with areal tendencies.
     [DG: The accusative ㄹ and the participial ㄹ differ in that the latter hardens the initial stop of the next word if applicable. It seems very unlikely that a flipped or rolled `r' could build up enough pressure to harden a stop; the obstruent `l' seems much more up to the task. This agrees with the author's analysis.]
[DG: This seems to be confirmed by Hulbert's nä-ga-kal-i-ta `I am the one who will go'. Hulbert (1905) normally distinguishes l and r.]

* Vovin, A., Interpretation of the Hüis Tolgoi Inscription, 2017, pp. 12.
Morphological analysis and interpretation of the second stone of the Hüis Tolgoi Inscriptions, using the transliteration by Maue. The language is clearly related to Mongolic (more than to Khitan). This brings the oldest record of a Mongolic language back to around 600 AZ. [DG: I am again confused. The author calls the language Ruan-ruan, in contradiction to his earlier publications.] The morphemes are largely recognizable as Mongolic, but there are considerable lexical differences, which makes translation difficult. The text praises the qaγan, calling him "the Qaγan of the Inscriptions", but there is not enough material to produce a coherent translation.
     There is another stone, the Bugut inscription, in the same script but perhaps a different language.

* Maue, D., The Khüis Tolgoi Inscription: Signs and Sounds, 2017, pp. 29.
Detailed description of the variant of the Brāhmī script used in the Khüis Tolgoi inscription, transliteration, and a preliminary transliteration, all on the assumption that the language is Mongolic-like.

* de Mol-van Valen, T., A Comparative Study on the Sayan Languages, (thesis), 2017, pp. 143.
Describes, among many other things, aspirated vowels in Dukha (Turkic) (p. 64).

* Ünal, O., On the Language of the Argippaei: An Ancient Predecessor of Mongolic?, 2017, pp. 5.
It is suggested that Herodotus' ἄσχυ may be a precursor of Written Mongolian *esüg `sour beverage' (with advanced vowels). More hesitantly it is pointed out that the name of the tribe, Ἀριµασποί, which Herodotus explains as ἄριµα ‘one’ and σποῦ ‘eye’, in a different segmentation matches *erim isepü `having maimed eyes', again with advanced vowels. [DG: This is information from about 450 BZ, and Herodotus was already reporting older data! If this is correct, Mongolian was already a recognizable language some 2500 years ago.]

* Janhunen, J., The Turkic Plural in -s, 2017, pp. 14.
There is ample evidence for a petrified plural marker *-s in Turkish: body parts that come in pairs, and the plurals of the 1st and 2nd pronouns: biz and siz, to name two. With rhotacism the marker presents itself as -r in Bulghar Turkish. The plural -s is also found on Mongolic, but it cannot be borrowed from Turkish, because Mongolic was only in contact with Bulghar Turkish, where is was -r. In Tungusic there is a suffix *-sA, which creates collectives. Whether these are accidental similarities or very early borrowing is unclear. The author does not consider them genetically related without further proof.

* Georg, S., Altaic Languages, in Encyclopedia of Chinese Language and Linguistics, 2017, pp. 7.
Although much of the text is, naturally, of reference book level, it also contains a number of detailed facts.
* The Altaic character of Mongolian is recent, developing in the 13-th century; likewise the more remote Tungusic languages are far less Altaic structurally than Manchu.
• Although there is no dearth of material trying to prove the genetic relationship hypothesis of the Altaic languages, there is far less material researching the areal relationship hypothesis.
• Modern Uyghur is not the descendant of the Old Uyghur of the Buddhist texts, but a variant of Uzbek. The modern descendant of Old Uyghur is Yughur (16k people).
     There is an older version of this article in another encyclopedia (`Encyclopedia of Language and Linguistics' (2005)?), with a table of the `Earliest Independent Attestations of Selected (Micro-)Altaic Languages', and a map of the Tungusic languages found in the Lower Amur region.

* Robbeets, M., Farming/Language Dispersal: Food for Thought, 2017, pp. 23.
This is Chapter 1 of `Language Dispersal Beyond Farming', edited by M. Robbeets and A. Savelyev (2017).
     It is traditionally assumed that languages spread with farming, and that the two go hand in hand. The book describes 15 centers of language spread, each showing that reality is more complicated. For example the Transeurasian languages spread from the homeland in the West Liao River Basin (± 5700 BZ) by agriculture. Part of the population reached East Mongolia around 1200 BZ (a distance of perhaps 600 km in 4500 years). By that time they spoke proto-Turkic (which makes East Mongolia the homeland of the Turkic languages). Then they became (horse-riding) pastoralists, and began their trek west, which ended in Turkey in 1071 AZ (a distance of at least 6000 km in about 2300 years).
     Changing a culture from subsistence to agriculture must take centuries if not millennia, so a binary statement as to whether a culture has agriculture or not is not tenable. Many factors contribute to language dispersal: technological advances, climate change, geographical conditions, war, etc., and agriculture is one of those factors.

* Georg, S., Other Isolated Languages of Asia, 2017, pp. 23.
Exhaustive enumeration and treatment (in variable depth) of the non-major languages of Asia, divided into two groups, North Asia and South Asia. [DG: But where is Burushaski? Ah, it is not "other" enough.]
     North Asia: Ket (6 pages, with a reasonable extensive explanation of the verb forms); Nivkh (5 pages, main features); Yukaghir (2 pages, showing that Kolyma Yukaghir and Tundra Yukaghir share structure and morphology but not the lexicon); Chukchi-Kamchatkan + Itel'men (1 paragraph); Korean and Japanese (1 page, explaining their status as isolates). [DG: Here we could repeat the phrase "share structure and morphology but not the lexicon".]
     South Asia: Kusunda (2 pages, with sample sentences and conjugation tables); Nahali (2 pages; history, noun declination, sample of its native lexicon (25% of total)); Great Andamanese (1 page; actually the Great Andamanese language and the Ang or Ongan languages; these two share structure and morphology but not the lexicon); and there is the mysterious and unapproachable Sentinelese.
     [DG: It is striking that the phrase "share structure and morphology but not the lexicon" turns up three times here. It might be a typological feature of language pairs that live next to each other.]

* Robbeets, M., The Transeurasian Languages, 2017, pp. 41.
A list of 27 typological features that set the Transeurasian (the author's term for macro-Altaic) languages off from their direct neighbours, based on the main languages of the 5 branches (Evenki for Tungusic) plus the oldest recorded language for each branch, for a total of 10 languages.
     Examples of macro-Altaic features are:
F2: Absence of complex tonal distinctions.
F4: Presence of tongue root vowel harmony.
F12: Property words may be verbally or nominally encoded. This means that adjectives (author's term: `property words') do not seem to form a class by themselves: some adjectives act like nouns, others like verbs, and some deviate totally. For example, in Korean many adjectives are verbs: nop- `to be high'; perhaps equally many use a word from an unnamed class + the verb ha- `to do/be': pigon ha- `to be tired'; and a very small number is immutable: heot `spoiled' (which is why in Korea one has to order hot coffee as hat). This "feature" is not found in any of the surrounding languages.
[DG: This seems to me one of the stronger indications that the macro-Altaic languages are in the end genetically related. This defect is not something languages will borrow, nor will it arise spontaneously in so many languages.]
F15: Morphology is agglutinative.
F20: SOV (Subject-Object-Verb) sentence order.
F22: Extensive use of converbs. Although converb constructions are possible in many non-macro-Altaic languages (English: "Having donned his coat he left.") they are unusual, whereas they are the normal way of expression in the macro-Altaic languages.
     The author gathers some interesting statistics from the matches between languages and features; some examples are:
1. The lowest number of matches any macro-Altaic language has is 19 (Modern Japanese), and highest number of matches any surrounding language has is 17 (Yukaghir); so the 27 features are really macro-Altaic-specific.
2. The language with the highest match quota is Middle Mongolian, with 26 matches out of 27: it is the "most macro-Altaic" language (missing feature: F9 Preference for a non-verbal strategy with (extra-family) verbal borrowing, i.e. borrowed verbs are integrated into the morphology in Middle Mongolian, rather than used with an auxiliary verb, e.g. `to do' or `to be').
3. The proto-versions have about 10% more matches than their modern successors, which suggest that still older versions would be still more "macro-Altaic".
4. Modern Turkish (20 matches), Modern Korean, and Modern Japanese (both 19 matches) have 4 to 5 fewer matches than the rest, corresponding to their geographic position, and the centrality of Mongolian.
[DG: More statistics: 13 of the 27 features apply to all 10 languages; 3 apply to 9 languages; and 1 applies to just 3 languages: F11 Inclusive/exclusive distinction in first person plural pronouns, only found in Old Mongolian, Manchu and Ewenki.]

* Bomhard, A.R., A Sketch of Proto-Altaic Phonology, 2017, pp. 18.
After sketching the history of the study of Altaic, the author shows and discusses Poppe's (1960) proto-Altaic phonological system, with 4 positions for stops (voiceless, voiced, and nasal), s, l and , r and , and 9 vowels (short and long). The present-day version is taken from the EDAL (2003) and has 4 positions for stops (aspirated, voiceless, voiced, and nasal), s, š, and word-initial z-, l and ĺ, r and ŕ, and 5 vowels (short and long). The structure of pronouns and particles was *(C)V, and other word stems were typically *(C)VCV or *(C)VCCV, with a strict distinction between nominal and verbal stems.
     On the basis of this the author expresses the opinion that Turkic, Mongolic, Tungusic, and perhaps Koreanic need not be considered separate branches of proto-Nostratic: the features they have in common "may be looked upon as due to common descent from an Altaic parent language"; on the other hand, Japanese consists of Altaic elements superimposed on an Austronesian language. The hypothesis that Japonic would belong with Korean or Altaic needs much work before it can be accepted.
     On the last page the consonants correspondence table from Starostin—Dybo—Mudrak (2003) is presented, which the author considers state of the art. It includes Korean and Japanese, in spite of that author's misgivings about their relation to Altaic. As to be expected the table is fairly tame (see Georg, `The Poverty of Altaicism' (2015)). The most complicated correspondence is

    PA *ŋ- → Mong. *Ø- / *y- / *g[u]- / *n[a o e]
which sounds reasonable.

* Robbeets, M., Austronesian Influence and Transeurasian Ancestry in Japanese: A Case of Farming/Language Dispersal, 2017, pp. 42.


* Francis-Ratte, A.T., Proto-Korean-Japanese: A New Reconstruction of the Common Origin of the Japanese and Korean Languages, (thesis), 2016, pp. 500.
Lots and lots of well-thought-out material.
     Chapter 2 explains the methodology and addresses some criticism to earlier proposals. Chapter 3 discusses phonological correspondences and processes, again with an eye to criticism to earlier proposals. [DG: It is striking that the reconstructed Proto-Korean-Japanese is much more similar to Korean than to Japanese.] Chapter 4 supplies 14 verbal morphemes and 7 nominal morphemes, connecting, e.g., MK tolh with OJ tati, both plural markers. Chapter 5 supplies 434 core lexical correspondences, most with pretty long explanations, which are needed to connect e.g. MK thwoski to OJ usagi, both `rabbit', or MK ithul ‘two days’ to OJ ito ‘thread’. Also interesting are MK twulh ‘2’ ~ OJ towo ‘10’, and MK cumun ‘thousand’ ~ OJ ti ‘thousand', and many others. This chapter includes a special section explaining the Korean numerals and connecting them to Japanese, but it is heavy going.
     [DG: This all feels to me like trying to prove that Old Irish and Armenian are genetically related, without the benefit of the rest of Indo-European. Just consider the numerals:

	Old Irish    oin dau   tri   cethir coic se   seyth ocht(n) noi(n) deich(n)
	Armenian     mek yerku yerek chors  hing vets yot   ut      inne   tas
I am not sure there is enough information left in both languages to prove their relationship. Of course, this does not prove that Korean and Japanese, because they are in a similar situation, are related.]
     Chapter 6 is concerned with distinguishing between borrowings and cognates, basically emphasizing again that borrowings betray themselves by irregular sound correspondences. It is also pointed out that there were no long periods of contact allowing large-scale borrowing [DG: Americans in Korea have shown that 30 years is enough].
     Chapter 7 summarizes the thesis. In a last section the author points out that Proto-Korean-Japanese had words for dry rice cultivation but not for wet rice cultivation, narrowing the break-up point to between 2500 BZ and 1500 BZ.

* Yondon, L., Japanisch und Mongolisch: Sprachvergleichende Untersuchungen, (thesis), (in German: Japanese and Mongolian: Comparative Investigations), 2016, pp. 226.
In this PhD thesis, the author compares a large number of features of Modern Japanese to the corresponding features of Modern Mongolian. The comparison also includes writing systems (the older systems are mentioned but no examples are shown), sound imitations, honorary forms, idiom, and calligraphy.
     Few fundamental differences were found: negation is expressed differently; auxiliary verbs behave differently; postpositions govern cases in Mongolian and are just postpositioned in Japanese; and there are differences in the use of honorifics.

* Shagal, K., Relative Clauses in the Languages of Sakhalin as an Areal Feature, 2016, pp. 18.
There are four languages native to Sakhalin: the Tungusic languages Uilta, Nanai, and Ewenki, and the isolate Nivkh. Their original word order was very probably SOV, but under influence of Russian the Tungusic languages have switched to SVO. The standard way of forming relative clauses (and in many cases subordinate clauses as well) was by using participles. The participles can be modified with nouns in various cases and adverbs to varying degrees in the four languages, and in some languages there is concord between the participle and the noun it modifies. These differences are analyzed and examples are shown. Due to its polysynthetic nature Nivkh is more resilient to absorbing Russian language constructions than the Tungusic languages. The author suspects that many speakers of Nivkh switched to Russian altogether rather than absorbing Russian elements.

* Blažek, V., Schwarz, M., Numerals in Mongolic and Tungusic Languages with Notes to Code-Switching, 2016, pp. 26.
After pointing out deep genetic relationships for the numbers 1, 2, 4, 5, and 10/100, and showing influences from various sources, the authors provide detailed tables of the units and teens in 21 Mongolian and 27 Tungusic languages and dialects. Several dialogues using borrowed numerals are reported.
     The authors see the present situation as the result of divergence from a common source, the Altaic language continuum, for about 6 millennia, followed by convergence for the last 2 millennia.
     Diagrams of language disintegration, with time scales, for Altaic (starting at 6000 BZ(!)), Mongolic (starting at 910 AZ), and Tungusic (starting at 800 BZ) close the paper.

* Alonso de la Fuente, J.A., review of Robbeets `Diachrony of verb morphology: Japanese and the Transeurasian languages', 2016, pp. 8.
Critical review of same, the main complaint being that the book hardly brings anything new. There are many detailed criticisms, for example:
• Sometimes a more close-to-home etymology is ignored in preference of a more exotic one.
• Many morphemes can be recognized as grammaticalizations of particles. This means that these morphemes are (relatively) recent, and thus have little probative value. Also, similar grammaticalization processes may occur in unrelated languages.
• The sound changes are trivially one-to-one (except a few that handle palatalization), which again points at the (relatively) short time depth of the comparisons. This makes it unlikely that the proto-Altaic level has been reached.
     There is also more general criticism:
• Use has not been made of the oldest Manchu writings.
• It is claimed that the results of the author's work on Altaic are "comparable in quality with those in the field of Indo-European studies". But this ignores the heavy criticism that this work has received.
• The romanization used for Manchu is confusing: The second back vowel of Manchu is traditionally denoted û. The author uses Gorelova's romanization, which uses ū, but transcribes this as u:, leading the reader to believe that it is a long u. But Manchu has no vowel length distinction.
     In summary the book is an up-to-date presentation of the pro-Altaicist view of verb morphology.

* Robbeets, M., Insubordination and the Establishment of Genealogical Relationship Across Eurasia, 2016, pp. 38.
Very detailed analysis of two proto-macro-Altaic deverbal noun suffixes: a neutral suffix *-rA and a resultative suffix *-xA. The first tends to develop an "uncertainty / non-past" meaning, the second a "finite past" meaning. Both suffixes are used in direct desubordination (i.e. as is, without any other syntactic material) to form main sentences. The author argues, supported by many examples, that the desubordination process is similar in the 5 branches of macro-Altaic. This is indicative of a genetic relationship between the macro-Altaic languages. The results are summarized in many tables.
     [DG: In Section 3.2 there is the enigmatic statement "The relative clauses marked with pK *-(wo)-l developed one step further to mark syntactically independent sentences, which added supplementary information in discourse, such as question, exclamation, confirmation and explanation.", cleaimed without literature reference. There is only one sentence ender in Korean that ends in -l, `-gol' and it stands out as anomalous in Korean grammar. It designates a mildly surprising or adverse circumstance. Does the author refer to this ending? It is certainly not a general development as the author implies.]
     There are no examples of the suffix *-xA in the Korean branch, which makes borrowing unlikely. [DG: The geographical position of Korean and Japanese was reversed for millennia; so that argument is weak.]

* Vovin, A., Vajda, E., de la Vaissière, É., Who Were the Kjet and What Language Did They Speak?, 2016, pp. 20.
[DG: Names are confusing here. The way I understand it is: Hsiung = Xiong = Hiung, and Chieh = Jié; the latter is one of the languages the Xiong-Nu spoke, and the only one we have information on.]
     The paper by Shimunek et al. `The Earliest Attested Turkic ...' (2015) claims that the Xiong-Nu spoke a Turkic language, but the author shows that the paper suffers from anachronisms and incorrect representation of Pulleyblank `The Consonantal System ...' (1962).
     The ethnonym 羯, which was pronounced as *jié in Middle Chinese (formerly written Chieh), and which was given the Late Old Chinese pronunciation of *kɨr by Shimunek et al., is reconstructed in Old Chinese pronunciation as *kjet by the author. This suggests an association with Yenisei people.
     The author now turns to the phonetic values Shimunek et al. assign to the characters of the Jié/Kjet couplet "Army go-out, <title> capture", and points out numerous errors: use of outdated information, incorrect quoting of standard material, and inconsistent quoting of the same source. Another reason the author rejects the resulting pronunciation of the couplet is that it does not rhyme.
     After chiding Shimunek et al. for taking liberties with the embedding Chinese text, the author spends the next 6 pages to discredit their Turkic interpretation of the couplet "in the light of multiple mistakes, mis-citations, wrong definitions and analyses, [and] forced and/or teleological constructions", each of which is explained in excruciating detail.
     The Yeniseian languages consist of two dialect/language groups, with Ket and Yugh in one group and Kott and Pumpokol in the other. [DG: The examples in the paper show that both groups differ far more than English and German.] A characterizing difference between them is that in the Kott group the subject particle in verb forms is a suffix and in the Ket group it is the first prefix, whereas in adjective forms it is a suffix in both groups. Using the most modern data on the pronunciation of Middle Chinese and on the Yeniseian languages, the author identifies the two verb forms in the couplet as the Pumpokol words t-i[lr]-ek-aŋ `[they] went out' and kat-a-qot-aŋ `[they] will catch'.
     The identification of the language as related to Pumpokol tallies with the many local river names of Pumpokol origin.

* Georg, S., The Poverty of Altaicism, 2015, pp. 37.
The paper consists of two parts: a short history of Altaic, largely taken from the author's `Japanese, the Altaic Theory, and the Limits of Language Classification' (2003) and the script of a speech addressing Martine Robbeets.
     In the speech, the author points out that the Altaic hypothesis is weak because it has "gaps". One such gap is that Mongolic words starting with `m-' never have cognates in Turkic. A simple explanation could be: "Of course there aren't, Turkic allows initial nasals only in exceptional circumstances." But if Turkic and Mongolic stem from the same proto-source, where did the proto-words that later started with `m-' in Mongolic, go in Turkic? A better explanation would be that at some point in time Mongolic imported an amount of Turkic words, which explains the observed similarities. These words never started with `m-', because Turkic does not have such words. The `m-'-words in Mongolic stem from proto-Mongolic which was disjoint from Proto-Turkic.
     Another problem is that the EDAL uses vowel correspondences of such "generous latitude" that "the vowel systems of any two languages might be made to seem related".
     In response To Robbeets' "Can Verbal Morphology End the Controversy?" the author presents a list of 17 Altaic morphemes from Robbeets' publications, with examples from Turkic, Mongolic, and Tungusic languages, only to present the audience with examples of the same 17 morphemes from Mari, an Uralic language. Does that mean that Mari is an Altaic language? That Ural-Altaic exists? That Nostratic exists? The author then shows that the detailed semantics of the similar morphemes is actually quite different.
     The crucial difference between Uralic and Altaic linguistics is that Uralic linguistics got results and consensus in a few decades, and that Altaic linguistics has obtained little result and no consensus in 150 years.

* Ulman, V., The Language of the Koguryŏ State: A Critical Reexamination, 2015, pp. 16.
Reexamines the etymologies and conclusions of Beckwith (2004) in the light of new pronunciations of 12th century Chinese (Baxter et al., `Old Chinese: A New Reconstruction', 2014), Some of Beckwith's etymologies are invalidated, but basically his conclusions stand.

* Gusev, V., Some Parallels in Grammar between Nivkh and Tungusic Languages, 2015, pp. 14.
Two morphemes and a particle are presented that are shared between some Tungusic language (mainly Negidal) and some dialect of Nivkh. One morpheme is a straight loan from Negidal to Nivkh, where it fills a gap with peculiar semantics: a 1st person imperative. The other morpheme has different forms in both languages, and has become more similar in use due to shared shift. The particle, gune in Negidal and furu in Nivkh, is used at the end of a sentence to indicate hearsay. The Negidal particle derives from Negidal gun- `to say' and the Nivkh one from Nivkh p‘ur-/fur- with the same meaning. They underwent shared development, helped by their superficial similarity.

* Vovin, A., Korean as a Paleosiberian Language, 2015, pp. 23.
Points out several typological correspondences between Korean an the Paleosiberian languages, but then backtracks and concludes "Korean is not a Paleosiberian language", because
1. typological correspondence is nowhere proof of relationship;
2. Paleosiberian languages are not a phylum, so you can't "belong" to them.
     [DG: In the meantime the indicated correspondences may well point to convergence processes which can tell us something about the early history of Korean. This is ignored here.]

* Janhunen, J., On Para-Mongolic vs. Pre-Proto-Mongolic Loanwords in Jurchen-Manchu, 2015, pp. 6.
Jurchen-Manchu derives its non-Tungusic look largely from a number of Mongolic-looking words. These words can be loans from proto-Mongolian itself or from one or more para-Mongolic languages. Only one para-Mongolic language is known: Khitan, spoken to the west of Jurchen in Southwest Manchuria. Lately more and more of the language has been deciphered, yielding more words and new opportunities for comparison.
     The preliminary results are:
• The colour terms in J/M cannot have come from Khitan, because the Khitan terms are completely different.
• The words for the numbers 11 to 19 in J/M probably do not derive from Khitan, because logographic forms in Khitan suggest they were expressed analytically (e.g. 11=10+1).
• For many other Mongolic-like words in J-M the evidence is negative (no, it is not from Khitan), or inconclusive.

* Kassian, A., Zhivlov, M., Starostin, G., Proto-Indo-European-Uralic Comparison from the Probabilistic Point of View, 2015, pp. 47.
The probabilistic method described by Baxter & Manaster Ramer (1999) was applied to a list of 50 words from proto-IE and proto-Uralic, representing concepts selected according to strict criteria, using GLD (Global Lexicostatistical Database) consonant classes. This resulted in 7 matches, which after calibration with the permutation test amounts to p = 1.9%, which is below the 5% limit but above the 1% limit. Rerunning the tests with more precise consonant classes in which the S-class has been split in sibilants and affricates, and L and R are distinguished, yields the same 7 matches, and p = 0.5%. So there are statistically strong indications that the chosen list is not random. The test does not tell what the cause(s) of the correlation could be.
     The 50 words are presented in an Appendix and it is shown that they all have extensive etymologies in the respective languages, which prevents them from being loans.

* Ringe, D., Response to Kassian et al., `Proto-IndoEuropean-Uralic Comparison from the Probabilistic Point of View', 2015, pp. 9.
Critique of the above paper of Kassian & al.
The author objects to the use of Dolgopolsky classes (which seem to be the same as the GLD classes above), pointing out that many sound correspondences in real-world etymologies often transgress the boundaries of these classes.
     The list is criticized for missing several highly stable words like `person', `fish', and `skin'. The PIE reconstructions do not contain laryngeals, but do use θ. [DG: Yes, the PIE etymologies are Russian-oriented.] Three of the 7 matches Kassian & al. obtain are mono-consonantal forms meaning `me', `thee', and `who', which make easy matches.
     Next the author argues that Kassian & al. show that there is a chance of 1:200 (p = 0.5%) that their list is random, but there are about 300 phyla, so it is no miracle that one among them matches. Kassian & al. should require 1:30,000. [DG: This would be true if Kassian & al. had set out to compare PIE to the 300 phyla of this world, but that is not what they did: they stated their hypothesis in advance, as one should in statistics.] Therefore the author rejects the result of Kassian & al., and even advices colleagues not to waste their time on the subject.

* Kessler, B., Response to Kassian et al., `Proto-IndoEuropean-Uralic Comparison from the Probabilistic Point of View', 2015, pp. 11.
Critique of the above paper of Kassian & al.
The author replicated Kassian & al.'s computations and obtained p = 0.3%.
     Emphasis is given to the influence of bias in the data, conscious or unconscious. This bias can also result from "imbalances in the literature and available data sets that are out of the researchers’ control". The Swadesh list was certainly constructed without bias towards Indo-European or Uralic, but it was not based on research in stability either. The choices made for the translations in both languages are a huge source of bias, and so is the choice of the reconstruction theory (traditional, laryngeals, glottals).
     The list used contained too many words that could be onomatopoeia (or "glug" words, as the author calls them). The list contained both lexical and grammar words.
     The comparison function used by Kassian & al. looks at both consonants of a root, but the second consonant is often much more garbled than the first, reducing the number of matches.

* Kallio, P., Nugae Indo-Uralicae, 2015, pp. 9.
Caustic critique of the above paper by Kassian & al.
Rather than comparing two proto-languages, Kassian & al. compare a 50-words list, and the author fails to see how this laziness would be progress. Kassian & al., true to the quip that in etymology vowels mean nothing and consonants very little, ignore the vowels and reduce the consonants to classes.
     Three of the matches are short forms and the others look much less convincing when laryngeals are inserted. [DG: This means that they could be loans from a time when the laryngeals were lost.] When we reject these 7 matches on the above grounds, Kassian & al. have shown that PIE and proto-Uralic have even less in common than was previously thought.
     The best evidence for Indo-Uralic lies in the paradigms, not in the lexicon.

* Kempf, B., Studies in Mongolic Historical Morphological Linguistics, 2015, pp. 235.
Considering the dictum that using the comparative method it is possible to reconstruct a language about 10 centuries back into past, the author surmises that starting from 13th century Middle Mongolian it should be possible to reconstruct 3rd century Mongolian. Such a reconstruction would be a large contribution to clearing up the question of the Altaic genetic unity.
[DG: I don't believe the dictum. It is blatantly impossible to reconstruct 10th century English from today's English alone. It might be possible given all the directly related languages: all Dutch, Frisian, and German languages and dialects, but no related languages are available for Middle Mongolian (except perhaps a few words of Ruan-Ruan).]
     As a first step to such a reconstruction the author analyzes all verb forms in the `Secret History of the Mongols' (Mongqol-un niuča tobča’an, or `Mnt' for short), and meticulously reconstructs 37 derivational verb-forming morphemes from them, 15 of which are deverbal verb suffixes and 22 denominal ones. The book closes with 25 pages of literature references and an index of verb forms.

* Kassian, A., Zhivlov, M., Starostin, G., Lexicostatistics, Probability, and Other Matters, 2015, pp. 17.
In which the authors address the points brought forward by the above three reviewers.
     The list of concepts was derived from that of the `Tower of Babel' project and based on 746 languages. The ASJP (Automated Similarity Judgment Program) proposed by Ringe is based on 245 languages, and was rejected because of the insufficient quality of some entries.
     The list was not biased. It is a standard list in which `louse' has been replaced by `liver' because no reliable PIE reconstruction for `louse' is available.
     It is agreed that the consonant classes are an approximation; they constitute the simplest of sound laws. But sound laws like those required by the infamous Old Indic dva- ~ Armenian erku `2' are seldom needed.
     The words for `nose' may be suspected of being onomatopoeia, but where does it stop? Is English [fut] an onomatopoeia because it sounds like stamping?
     The lists contain short words like `I' and `who', because they are integral part of doing etymology, and it would be funny to linguistic research excluding them. Also such words are often important in showing genetic relationship.
     The consonant classes are not biased. They were composed independently, without reference to PIE-Uralic.
     The 300 phyla argument is countered by saying that it only holds if one is indeed comparing PIE to 300 phyla.
     Whether laryngeals are used or not makes no difference, since they coincide with the absent consonant in the consonant classes.

* Robbeets, M., Diachrony of Verb Morphology: Japanese and the Transeurasian Languages, Mouton-De Gruyter, Berlin, 2015, pp. 569.
This book deals with shared verb morphology in Japanese and other languages that have been identified as Transeurasian (traditionally: "Altaic") in previous research. It analyzes shared etymologies and reconstructed grammaticalizations with the goal to provide evidence for the genealogical relatedness of these languages.

* Werner, H., review of Georg's `A Descriptive Grammar of Ket (Yenisei-Ostyak)', 2015, pp. 9.
High praise for this book. Half of the book is justifiedly dedicated to "His Majesty the Verb". The structure of the Ket verb form is similar to that in the Caucasian and Na-Dené languages, and has 10 positions:
     8 - 7(r) - 6 - 5(d) - 4 - 3 - 2 - 1 - 0 (R) - −1
The root part (R) has a general meaning, which is then narrowed down by (r), while the single-letter determinative (d) may say something about the circumstances.

* Shimunek, A., Beckwith, C.I., et al., The Earliest Attested Turkic Language, (The Chieh 羯 (*Kɨr) Language of the Fourth Century A.D.), 2015, pp. 9.
Starting from the same data as Vovin `Did the Xiong-Nu speak ...' (2000), and with equally convincing arguments, the authors reconstruct the phonetics of the text as

    *su-Ø    kete-r    erkan boklug-gu  tukta-ŋ,
interpret it as Turkic:
    army-NOM leave-AOR TEMP  boklug-ACC capture-2PL.IMP,
and give the translation "When/as the army goes out, capture the Boklug!".

* Vovin, A., Out of Southern China?, (slides), 2014, pp. 55.
[DG: These are slides, so the narrative is missing.]
     First the author investigates the claim that Japanese is an (almost) isolating language, refuting it by producing no less than 30 verb endings. These are then reduced to 6 basic endings, reinstituting the claim [DG: that is the narrative, isn't it?].
     Next, Janhunen's hypothesis that Japonic was originally monosyllabic and isolating and of a Sinitic type, is examined. The monosyllabic part is refuted by showing that many of Janhunen's etymologies of Japanese disyllablic words as compounds of two monosyllabic words are untenable.
     As step 3 lexical parallels between Japonic and Tai-K(r)adai are investigated, supported by a list of 23 etymologies of monosyllabic words (several of them basic words:, e.g. `wife', `water', `fire', `tooth'), with accent in Japonic and tone in Proto-Tai. It is striking that low accent in Japonic often (6 out of 8) corresponds to tone A2 in proto-Tai. Nine disyllablic etyma add more information.
     [DG: I don't understand the slides about Fangyan.]
     Benedict (1989) suggested that Japanese and Austric/Tai would be genetically related. The above points seem to confirm that, but there are two reasons to assume that the relationship is not genetic but rather originates from contact:
1. the pronominal system is not related;
2. there are almost no words with tone B in the etymologies (a "gap").
Details of the etymologies show that the contact must be very old and must have been very intensive.
     These conclusions clear up some questions in the Japonic languages; it suggest an Urheimat for them (Southern China); and it heralds the end of the Altaic hypothesis [DG: at any rate the Japanese-Altaic hypothesis; Japanese must have picked up its Altaic features much later, in Northern Asia].

* Unger, J.M., No Rush to Judgment: The Case against Japanese as an Isolate, 2014, pp. 20.
There are several arguments being made in this paper.
     After admitting that the Korean-Japanese hypothesis is not proven, the author claims that assuming it is false leads to even more unsatisfactory conclusions. Korean and Japanese are very similar, and similarity between languages can arise only from three effects: chance, borrowing, or genetic relationship. Korean and Japanese are too similar for just chance; historic facts show that Korean and Japanese were in contact over at most a couple of centuries, not enough to explain the strong similarity; and given that Francis-Ratte (2016) has shown a large number of common etymologies, the Korean-Japanese hypothesis is likely to be true, and Japanese is not an isolate.
     An important part of this paper is Appendix I, which contains a structured list of what we know about the history of Japanese and Korean, and Appendix II, which contains the corresponding maps. These facts are then combined into a narrative about the migrations that led to Korean and Japanese. [DG: All this is put on firmer ground in Unger (2019).]

* Robbeets, M., Common Denominal Verbalizers in the Transeurasian Languages: Borrowed or Inherited?, 2014, pp. 18.
[DG: On page 1 the author admits that `Transeurasian' "refer[s] to a large group of geographically adjacent languages, traditionally known as “Altaic” ". If it is that simple why is a new name necessary?...]
     The author observes that denominal verbalizers (morphemes that turn nouns into verbs) are more stable than inflectional markers (verb endings): e.g. French lost a large part of the Latin verb conjugation endings, but the Latin denominal verbalizer -sc- is still recoverable in e.g. nous fini-ss-ons `we stop'. Moreover, denominal verbalizers change the word class of the stem from noun to verb, which is a much bigger step than from verb to verb, which is what most verbal modifiers do. As a result denominal verbalizers are very resistant to borrowing.
     The author supplies six tests to recognize that a denominal verbalizer is borrowed rather than inherited. A simple example is the English denominal verbalizer -ify and the Latin denominal verbalizer -ificere. If they were both inherited from a common source, PIE, Grimm's law tells us that English f should correspond to Latin p, but Latin has an f. So English -ify must be borrowed from Latin rather than inherited from a common source. The other five criteria are more complicated. This is followed by five tests to recognize that a denominal verbalizer is inherited rather than borrowed. For example, if two languages share a grammaticalized morpheme and the components of the grammaticalization, the morpheme is almost certainly inherited rather than borrowed. French, Spanish, and Italian all share morphemes that reflect the Latin causative -ificere and still have a form of the Latin facere. So one can conclude that they inherited the morphemes.
     These tests are then applied to the Transeurasian languages, with the result that it is "safer" to attribute similarities between some verbal morphemes to inheritance than to borrowing.

* Robbeets, M., A Velar Fricative in proto-Transeurasian, 2014, pp. 26.
Analyzing the behaviour of the resultative morpheme *-kA- ~ *-gA- ~ *-xA- in many languages in the five phyla of Transeurasian, the author concludes that there was a velar sound in proto-TEA which has x as its representative in several languages. The author tacitly assumes that that sound was x.
     [DG: That seems unlikely: the sound is almost always represented by k, and fricatives rarely revert to plosives: it forces the author to use the term `de-fricativization'. Starostin reconstructs an aspirated k. Creaky voice on the vowel would also provide an explanation.]

* Schmidt, R.W., Seguchi, N., Jomon Culture and the Peopling of the Japanese Archipelago: Advancements in the Fields of Morphometrics and Ancient DNA, 2014, pp. 26.
This research "suggests a level of inter-regional heterogeneity not expected among Jomon groups," many hailing from Southeast Asia. The Siberian haplotypes found in the Ainu may be "recent", perhaps 5th century AZ. [DG: If I read correctly this material about which I know next to nothing.]

* Robbeets, M., The Japanese Inflectional Paradigm in a Transeurasian Perspective, 2014, pp. 36.

Inflection does not change the basic semantics of the root (`I count' vs `I counted') , whereas derivation does (`to count' vs `to discount'). The semantic changes in derivation are often quirky, and if the same quirk is found in an other language, it can contribute to a proof that the two languages are "linguistically related".
     Borrowed morphemes are often borrowed together with the lexical form and/or other morphemes, in a combination that makes sense in the original language, but no longer in the borrowing language. This provides a way to recognize them.
     Morphemes are typically very short, so accidental matches are likely. This is counteracted by the usually small number of morphemes in a language. Still coincidences occur and the author gives the famous PIE ~ proto-Eastern-Miwokan example.


* Kisamov, N., The Hunnic Oracle, 2014, pp. 15.
Largely based on Taskin, `Materials on the History of Nomadic Peoples in China' (1990).
     The author takes it as a fact that the Hiung-Nu were the Huns, and claims that there is consensus that they spoke a Turkic language. The idea that they could have spoken a Yeniseian language is rejected: 1. 17th century Yeniseian does not have the vocabulary for ruling a very large empire; 2. identifying ten syllables from an unknown language from the 4th century with languages known only from the 17th century would be utterly unsound. [DG: Can one from 10 syllables of the Beowulf (without translation) identify them as Old English? It seems I might, but perhaps I know too much. And also, the Beowulf is 3 centuries younger than the Hunnic oracle couplet]. Mongolic languages are not considered.
     The author presents in a table the results of the previous attempts at interpreting the couplet. These are mostly based on Karlgren's phonetic reconstructions. Returning to something close to Ramstedt's transliteration, Taskin romanizes the Hunnic couplet (which is written in Chinese characters as: ⚽支 替戾剛, 僕谷劬 ⚼當) as
"Xiù-zhī tì-lì-gāng, púgǔ-qú tū-dāng". Taskin claims that this is almost immediately recognizable as the following sentence in old-fashioned Turkish:
"Süči tiligan, Pugu'yu tutan", which means "(If) army-man (would) go, Pugu would (be) capture(ed)".
Detailed explanation of the suffixes is given: -či - standard nominal agent affix; -gan - past participle, 3p., perfect tense verbal affix; -'yu - future conditional verbal transitive affix; and tutan is explained as: “capture” in 2nd or 3rd p. sing. future tense ablative ending, passive voice.
     It is also significant that the verse shows a five-syllable metre, typical of ancient Turkish poetry.

* Ko Seongyeon, Comparative Consequences of the Tongue Root Harmony Analysis for proto-Tungusic, proto-Mongolic, and proto-Korean, 2014, pp. 36.
Many of the arguments in this paper can also be found in Joseph `Comparative Approach of the Vowel ...' (2020), often presented in a more systematic way.
     Based on extensive analysis of the tongue root positions involved in the vowel harmony systems of several North-East Asian languages, the following conclusions are drawn:
0. Vowel harmony can be based on one of four features: [±front] (palatal); [±round] (labial); [±high]; and Tongue Root Harmony, which can be Advanced [±ATR] or Retracted [±RTR].
1. Korean, Mongolic, Tungusic, Chukchi, and possibly Yukaghir have vowel harmony based on the feature [RTR], not [front] or [ATR] as has been claimed in the literature.
2. This [RTR]-based vowel harmony is an ancient North-East Asian areal phenomenon, not an innovation in some languages.
3. Kalmyk and Oirat lost their [RTR]-based vowel harmony (in favour of palatal vowel harmony) -- the other 15+ Mongolian languages did not acquire it independently.
4. The Turkic languages lost their [RTR]-based vowel harmony (in favour of labial vowel harmony) -- Mongolic, Tungusic, Korean, and Chukchi did not acquire it independently.
5. In as far as the above language families are genetically related, their parent language, proto-Altaic, had [RTR]-based vowel harmony.
6. [ATR]-based vowel harmony is an areal phenomenon in the Central Sudanic Zone, found in languages of the Niger-Congo, Nilo-Saharan, and Afroasiatic phyla.
[DG: Joseph `A Comparative Approach of the Vowel ...' (2020) has clear tables (Tables 29.1-5).]
     As Janhunen (`Korean Vowel System in ...' (1981)) pointed out, the vowel inventories of pre-Old Korean, proto-Mongolic, and proto-Tungusic are almost identical. The only difference lay in the two central back vowels, which were o and ɔ (both with [+round]) in proto-Mongolic and proto-Tungusic but ɨ and ʌ (both with [−round]) in Old Korean. (The four vowels were further distinguished by o and ɨ having [−RTR], and ɔ and ʌ having [+RTR]). This anomaly could be explained by the following scenario.
1. Pre-Old Korean had o, ɔ ([+round]), just as proto-Mongolic and proto-Tungusic.
2. There was a tendency in pre-Old Korean for these central vowels to lose their roundness.
3. That loss would, however, cause the differences between o and ə, and between ɔ and a to disappear.
4. This was compensated for by raising o and ɔ to ɨ and ʌ. This anomalous development might have contributed to the weak position of the ɨ in Modern Korean and the disappearance of the ʌ in Early Modern Korean.
[DG: In a situation in which raising an o resulted in an ɨ, raising an ɔ might well have resulted in an ɯ, as suggested by me in the annotation to Lee `Geschichte der koreanischen Sprache' (1977).]

* Stachowski, M., Protolanguage Models, Reconstruction Methodology, Khakas, Oghuz and English, 2014, pp. 12.

Le proto-Oghuz n'existe pas!
     Proto-Oghuz is considered the parent language of Modern Turkish, but that concept is shaky. First, Modern Turkish includes some Kipchak features (the author does not specify which ones), and Kipchak is not Oghuzic. Second, Turcology recognizes 12 features that set Oghuz apart from the other Turkic branches, but the author shows that these features arose in several separate dialects of Oghuz over a period of several centuries. So what was this proto-Oghuz language and who spoke it? The answer must be that it was a cluster of mutually intelligible speech forms, and no, nobody spoke proto-Oghuz the way it is in the books.
[DG: I suppose this is true for most "proto"-languages, except perhaps those of the most successful comparative reconstructions, in particular Indo-European. It is likely that there was never a proto-Altaic, just a bunch of dialects.]
     As other examples of single languages originating from clusters of dialects, the author points to Khakas and English. The Khakas languages did not exist until the 1920's, when Soviet authorities declared the Turkic tribes Beltir, Kača, Koybal, Kızıl and Sagay to be one "nation" and gave it the name Khakas; and almost a 100 years later there is a standard Khakas language, with literature and culture. Likewise Middle English does not descend from a single Old English dialect but is rather the result of convergence of Anglo-Saxon, Celtic, Scandinavian and Romance elements.
     Incidentally, Ramstedt wrote in 1903 words to the following effect: "When the North-Asian nomadic peoples, unfettered by borders or agriculture, would live and fight together over millennia, exchanging 99% of their vocabulary, would that not be a `proto-language'?".
[DG: I think Ramstedt misses the point here: his language would be quite homogeneous. But North-Asia is large and people were few and far between, there were no horses yet so people had to go everywhere on foot, so contact was rare, and every extended family had their own somewhat different family speech form. And Turkish resulted from the convergence of speech forms A, P, and X; Mongolian from B, Q, and Y, and Tungusic from C, R, and Z.]
     Khakas is not a creole, because a creole implies the prior existence of a pidgin, and none existed for Khakas. "Koine" would be a better term, as perhaps for proto-Oghuz. [DG: No. Koine is spoken and nobody ever spoke the fantom Oghuz, by the author's own claim.]

* Janhunen, J., Personal Pronouns in Core Altaic, 2013, pp. 16.
The 1st and 2nd person pronouns in all language groups in northern Eurasia, from English (`me'-`thee') to Yukagir (`met'-`tet') (with the exception of the Yeniseian languages) show signs of the pattern `m'-`T'. The author describes the situation but does not go into an explanation of this remarkable state of affairs.
     Within this (large) group, the three Altaic language groups stand out for two reasons: the 1ps starts with b rather than with m; and the sets of pronouns in these languages are similar beyond belief (see the table in the annotation of Vovin, `First and Second Person Singular Pronouns' below). The similarity extends beyond the first phoneme of the pronouns: they often have a trailing -n (or -m) in common (as does English `mine'). Initially (in the 19th century) this was taken as strong evidence of the genetic unity of the three Altaic languages, but hardly any other evidence of such a relationship was found. By all known rules of language development the three pronoun sets can hardly be a 1000 years old (or they would have diverged more; see the other table below), whereas, if the three languages have a common origin at all, that origin must lie at least 6000 years in the past (or reliable cognates would be found).
     Because `b'-`T' as an attractor state is inferior to `m'-`T', attractor action is ruled out as an explanation, and so is common origin (and coincidence, obviously). The bulk of the similarities found between Turkic and Mongolic, and Mongolic and Tungusic are explained through intensive language contact. Such a situation is favourable for `shared drift' to play a role (see Janhunen, `Non-Borrowed Non-Cognate Parallels in Bound Morphology' (2012)). The effect may be going on even in recent times: in three Mongolic languages the 2ps c has changed into s, making it more similar to the Tungusic 2ps.
     In short, the pronouns are not too new for Vovin's explanation to hold, aand they were well-preserved by shared drift.
     [DG: Vovin's explanation of borrowing from Bulghar Turkish some 2000 years ago, plus Janhunen's shared drift keeping them similar, explains for a large part (where does the b come from?) the uncanny similarity between the Core Altaic pronouns. But there is no proof that this was what happened, its only virtue is that it explains something that otherwise would be unexplainable. But then, we don't have proof of the correctness of quantum mechanics either; its only virtue is ...]

* Jankowski, H., Altaic Languages and Historical Contact, 2013, pp. 24.
In the first few pages the author dispels the misunderstanding that the Polish linguist Kotwicz adhered to the idea that the relation between the Altaic languages was genetic; on the contrary, in the 1930s "Kotwicz claimed that `many centuries BC' three groups of tribes which spoke similar, but distinct languages, Turkic, Mongolian and Tungusic, got close to each other in the territory of what is now Mongolia and north-east of it. The languages of these groups were differentiated and individual tribes started to separate quite early".
     The author's attempt to determine the home land of the Altaic languages fails, because all reasonable regions can be rejected on reasonable grounds. So the ball is with the adherents of the Altaic hypothesis, although the author suggests a "dynamically changing homeland". [DG: Wouldn't the earliest position of the d.c.h be the home land?]
     The author observes that many Altaic etymologies in older works (e.g. Ramstedt) are between specific pairs of languages: Korean-Tungusic, but far fewer etymologies are presented between Korean on one hand and Mongolic or Turkic on the other. Later etymologists (e.g. Starostin, Dybo) produce etymologies that cannot be explained phonetically.
     Proto-Altaic is supposed to have existed in the 4th millennium BZ, when culturally the most important words were `stone', and `clay'. Proto-Altaic `stone' is supported by Trk. tāš, Mong. čilagun, Tung. ǯolo [DG: + Kor. dolh?] but a borrowing chain Trk. tāš → Bulg. Trk. čul → Mong. čilaγun → Tung. ǯolo seems at least as acceptable, the more so because Tungusic has several other words for `stone', so ǯolo may be a late-comer. The common term for `clay' is also doubtful. Any etymologies for `bronze', `brass', `copper', or `iron' are anachronisms.
     Several examples of contact between Ancient Turkic, Ancient Mongol, and Ancient Tungus on one hand and Ancient Samoyedic on the other are given.

* Vovin, A., Northeastern and Central Asia: `Altaic' Linguistic History, 2013, pp. 7.
Concise description of same, reporting on modern research. The word `Altaic' is used in an areal sense; genetic relationships are recognized only when good etymologies are available.

* Vovin, A., The Mongolian Names for ‘Korea’ and ‘Korean’ and Their Significance for the History of the Korean Language, 2013, pp. 6.
After a condensed version of the arguments in Vovin (2017) the author turns to Middle Mongolian and Classic Mongolian, both of which differentiate between `l' and `r'. `Korea' was solγo in Classic Mongolian, and `Korean' solongγos. This is then identified as the Old Korean word selo, `Shilla'. This proves that the Old Korean word selo was pronounced with an `l'; and since the text was from 1228 AZ, the merging of `l' and `r' in Korean must have taken place after that date. [DG: That only works if there is evidence that some other word had `r' at that time.]

* Vovin, A., Why Koreanic Is Not Demonstrably Related to Tungusic, 2013, pp. 8.
After lamenting that refuting a bad etymology takes at least twice the space as the original etymology, the author notices
1. that most of the reported Korean-Tungusic etymologies involve Manchu or Jurchen, and almost never one of the more far-off Tungusic languages, which suggest mutual influence with a single language;
2. that all but 18 of the etymologies suffer from the usual sins: irregular phonetic matches, unaccounted segments, ghost words, borrowings from different languages, and semantic mismatch; five examples are given.
     The 18 "good" etymologies are then all shown to be better considered as loans (mostly from Middle Korean to Manchu) because the proposed correspondences do not explain the whole word [DG: this is an unusual strict requirement].

* Hatcher Jr, R.J., A Description of Korean Converbs and their Northeast Asian Context, (thesis), 2013, pp. 144.
This thesis strikes one for its high level of abstraction: it is very much "Tell, don't show", and sometimes it does not tell either. The subject is not introduced, and familiarity with Haspelmath is assumed. The difference between same-subject converbs and different-subject converbs is not mentioned until page 26, where it becomes relevant for explaining the semantics of the Korean -myense. That said, the thesis contains a lot of information.
     The 26 Korean converbs are first summarized in a table, and then presented one by one, specifying the semantics of each converb in fairly abstract terms, followed by one or two examples. This format is then repeated for Japanese; Ewenki, Udihe, and Manchu (Tungusic); Mongolian (Mongolic); Sakha (Turkic); and Nivkh, Ainu, Kolyma Yukaghir, and Chukchi (Proto-Asiatic).
     There are several conclusions:
• Some languages allow converbs to follow the finite verb, in spite of the SOV word order (called AOV here), perhaps under the influence of Russian.
• DS converbs are found to be rare [DG: but it is not clear whether "DS" means "different subjects are allowed" or "different subjects are required/implied"]. This may be related to the absence of subject markings in those languages.
• Many languages have a "favorite" converb, one with many more meanings and uses than the other converbs.
• Only Korean has converbs that can be used with verbs marked for tense, although some other languages allow marking for aspect on a converb.
• Converbs are not restricted to the North East Asian languages.

* Malchukov, A., Verbalization and Insubordination in Siberian Languages, 2013, pp. 34.
Although the paper is mainly concerned with scenarios for the origin of desubordinated forms in Siberian languages, it additionally shows that desubordination is an areal Siberian phenomenon rather than a specifically macro-Altaic one. This weakens Robbeets' argument in `Insubordination in Altaic' (2009).

* Ko Seongyeon, The End of the Korean Vowel Shift Controversy, 2013, pp. 30.
The notion of a Korean Vowel Shift arose from a discrepancy between the assumed sounds of the vowels of Old Mongolian and the way they were represented in Hangul in loans in Middle Korean. Revising the Old Mongolian and Middle Korean vowels sets by introducing the feature "Retracted Tongue Root" makes the discrepancy go away and with it the need for a Korean vowel shift.
     See Ko Seongyeon `Tongue Root Harmony and Vowel Contrast ...' (2012), and Ko Seongyeon `The End of the Korean Vowel Shift Controversy' (2011).

* Georg, S., Remarks on Anonymous Remarks on my `Poverty of Altaicism', 2013, pp. 7.
Saddening defense against an anonymous attack, on points that are only tangential to the gist of `Poverty of Altaicism'.
     The main gripes were that the author clarified Pallas' (1741–1811) term "die tatarische Sprache" as "the Tatar (= Chaghatay Turkic) language", and translated "Gemeinsamkeit beyder Nationen" as "symbiosis of both nations", for which the anonymous critic suggests "unity (oneness, commonality, similarity) of both nations", accusing the author of an attempt to mistranslate the text in order to "prove" that Pallas already doubted the Turkic-Mongolic unity.
     In answer to the first point, the author shows in detail that in Pallas' time "Tatar" meant any non-Ottoman Turkish language, and that Chaghatay Turkic was the only well-known one then. As to the second gripe, "Gemeinsamkeit" means `commonality'; `similarity' might be acceptable, `symbiosis' is wrong, and `unity' and `oneness' are very wrong. As to the accusation, Pallas' German text is very clear and leaves no room for the idea that he considered a Turkic-Mongolic unity.
[DG: I'm not sure I could have resisted the temptation to add an -e to `Anonymous'.]

* Johanson, L., Robbeets, M., Bound Morphology in Common: Copy or Cognate?, 2012, pp. 20.
Introductory chapter of Johanson & Robbeets `Copies versus Cognates in Bound Morphology' (2012), a collection of 21 papers reporting on attempts to find methods to distinguish between cognates and borrowings (copies) in bound morphemes. The paper consists of two parts. The first part is concerned with causes and effects of borrowing. The second consists of abstracts of the other 20 the chapters. Both parts are very information-dense.
     Four causes are identified for forms in different languages to look alike: 1. chance (proto-Miwok and proto-Indo-European have paradigm endings that are quite similar); 2. language universals (English and Chinese are both SVO); 3. genetic relationship (cognates), though genealogical proximity; 4. borrowing (puristically called "code copying" here), through areal proximity.
     When we restrict ourselves to bound morphemes, chance can be ruled out rather easily "by regularity and paradigmaticity" [DG: 1. what does that mean?; 2. both Indo-European and Korean use -i- for causativity/passivity]. But distinguishing cognates from loans is much more difficult. The last two decades have brought in more and more examples of pronouns, bound morphemes, and even entire paradigmata borrowed (no examples given). So it would seem that in the end anything can be borrowed. But that conclusion is too general.
     Some features are more easily borrowed than others: there is a borrowing hierarchy, basically lexicon > inflectional morphemes > derivational morphemes. Many refinements are discussed. Social factors can influence the above order. Ergativity is a feature that is seldom borrowed. And the striking presence of 1st an 2nd person pronouns with m-/T- in northern Eurasia requires a special theory: attractor states.
     There are many ways to spot borrowed morphemes from inherited ones; examples are:
• If a shared morpheme attaches only to shared roots (and not to other roots) chances are that both the roots and the morphemes are loans.
• If a bound morpheme has several meanings in language A and only one in language B it is likely that it is a loan from A to B.
• If a shared bound morpheme has more than one shared form on both language, it is likely that they are genetically related.
• If a shared morpheme occurs in just one dialect of one of the languages, it is likely that that dialect has borrowed it (but other people cite "fragmentary nature of textual evidence").
     Grammatically dissimilar languages can still borrow lexical items from each other, but borrowing morphemes is much less likely. Lack of segmentability of affixes is an obstacle to lending to an agglutinating language, e.g. preventing Basque from borrowing the fusional endings of Latin but allowing ample borrowing of Latin nouns. Morphemes with a clear-cut meaning are more easily borroweded than ones with a more fuzzy meaning.
     One-line summaries of the summaries of some of the chapters:
Ch.2: Janhunen about "shared drift"; see Janhunen (2012).
Ch. 3: Nichols about "m-T"; see Nichols (2012).
Ch. 4: Gardani about what gets borrowed and what not.
Ch. 18: Hayasi about the Eynu language, an Uyghur secret language that has replaced most of its lexicon by Persian words, for secret communication.
Ch. 19: Whaley about using derivational morphology to obtain insights in the Tungusic languages classification.
Ch. 20: Unger about Vovin's claim that any morpheme common to Korean and Japanese but not occurring in Ryukyu or Old East Japanese was borrowed from Korean by Western Old Japanese.
Ch. 21: Robbeets about guidelines to distinguish loan from cognate and the application of these guidelines to shared Altaic morphemes; see Robbeets (2012).
The other chapters concern non-North-Asiatic languages, mainly from the Americas.

* Robbeets, M., Shared Verb Morphology in the Transeurasian Languages: Copy or Cognate?, 2012, pp. 22.
The term "Transeurasian" is introduced as replacement of "(macro-)Altaic", without argument.
     Guidelines are set up to distinguish borrowings (the more correct but unusual name of "copies" is used here) from cognates = genealogical related items. These guidelines are then applied to a set of morphemes allegedly shared between the macro-Altaic languages.
     The following 6 guidelines are negative: they mark cases where inheritance is not likely.
1. Attachment of Shared Morphemes: If a shared morpheme attaches only to shared roots (and not to other roots) chances are that both these roots and the morphemes are loans. As an example a Turkish verb paradigm loaned to Romani is shown.
2. Shared Suffix Complexity: If a shared morpheme is meaningfully segmentable in one language but not in the other it was a loan from the first language. An example is a Yakut verb paradigm that is borrowed by two Tungusic languages. In this case the endings are obviously borrowings in the Tungusic languages, because they consist of two separate morphemes in Yakut, but are just opaque fused two-syllable endings in the Tungusic languages.
3. Shared Function: If a shared morpheme occurs in one language only in a meaning that is secondary in the other, it is a loan in the first language.
4. Contradiction of Established Sound Correspondences: If sound correspondences between the two languages are known and the shared morpheme does not obey these it is very likely a loan.
5. Areal Distribution: If a shared item only occurs in adjacent areas and not in more remote places, that is an indication against inheritance of the item.
6. Shared Paradigms: If a set of shared morphemes is restricted to one corner of the grammar, e.g. noun endings only, it is likely that they are borrowed rather than inherited.
     The following 6 guidelines are positive: they address cases where inheritance is likely.
7. Shared Infrequent Grammaticalization: French, Spanish, Portuguese, and Italian all form their future tense by grammaticalizing `infinitive + form of "to have"'. Since this specific pattern of forming the future tense is not a usual pattern (is "infrequent"), it is characteristic of the Romance languages and is indicative of inheritance.
8. Categorial Clarity: A morpheme is categorially clear if its grammatical function can be understood without considering its environment. A "categorially opaque" morpheme would mean one thing in the presence of one morpheme and another thing in the presence of another morpheme. Categorially opaque morphemes are never borrowed, so if one is shared, it is almost certainly inherited.
9. Shared Cumulative Function: If a shared morpheme has similar "cumulative functions" (e.g. 2nd person past tense passive) in both languages it is very likely that it is inherited.
10. Shared Variant Allomorphy: If the shared morphemes have more than one form in both languages and these forms match, this is a strong indication of inheritance.
11. Comparative Setting: Normally morphemes from two languages are compared. If more languages are involved, the likelihood of inheritance increases. [DG: I find this strange, and the given examples (e.g. Spanish (Indo-European) nominalizer -ero loaned to Nahuatl (Aztec), and from there to Sierra Popoluca (Mixe-Zoquean)) do not support this.]
12. Discontinuous Areal Distribution: If areas A, B, and C are adjacent in that order and an item occurs in areas A and C but not in B, it is likely to be genetically related in A and C, because it could not have been borrowed via B.
     The 12 guidelines above are then used to examine 23 alleged macro-Altaic morphemes, mostly briefly but sometimes in detail. The list of morphemes does not include the personal pronouns because 1. Nichols (2012) explains their similarity through attractor states rather than through inheritance, and 2. Old Turkic bän, bini ‘I, me' may come from b- `this, here' (Erdal (2004)), and thus may not be related to the very similar Mongolian and Tungusic pronouns at all.
     The conclusion is that none of the 23 morphemes fall under the negative guidelines 1-6 whereas many conform to the positive guidelines 7-12. [DG: It is indeed an impressive tableau.]

* Janhunen, J., Non-Borrowed Non-Cognate Parallels in Bound Morphology: Aspects of the Phenomenon of Shared Drift with Eurasian Examples, 2012, pp. 24.
In addition to divergence (drift) and convergence (borrowing) as causes of language change the author identifies `shared drift', in which language contact causes two or more languages to undergo similar changes in functional-formal elements that are otherwise unrelated.
     One example is the shared development of the conditional clause ending -sa, which in Turkic Sarygh Yughur derives from the verb sa- `to think' and in the neighbouring Mongolic Shira Yughur derives from the Old Mongolic conditional suffix -xAsA.
     This example could be dismissed as coincidence, but the author presents many more examples, including: details of the spread of the noun plural in -s in Western Europe; plural marking -s in the three branches of Altaic; dative markers in Mongolic and Tungusic; past tense marking -le[e] in Mongol and Mandarin; and the morphological lengthening of the final vowel in Helsinki Swedish and Helsinki Finnish.
     In the latter case the final -r of Swedish plural and 3ps is dropped and replaced by lengthening the preceding vowel. For H. Finnish the result matches the partitive ending (used as plural after numbers), and for the 3ps H. Finnish replaced the Standard Finnish ending -a by vowel lengthening, thus matching the H. Swedish long vowel for 3ps.
     These coinciding morphemes can easily create the illusion of a genetic relationship, so genetic research will have to look out for them.

* Nichols, J., Selection for m: T Pronominals in Eurasia, 2012, pp. 28.
Very many languages in Eurasia (synchronically) have a 1sg pronoun or morpheme starting with m and often a 2sg pronoun or morpheme starting with a `t'-like sound, summarized as `T' (the so-called `Mitian' languages). This `m:T' pattern can be found in the Indo-Uralic, Micro-Altaic, and Kartvelian languages, and in Yukagir. Diachronically it turns out that this pattern is sometimes the result of displacement of older pronouns. This paper examens the displacement mechanisms.
     The author considers the patterns `m' and `m:T' as attractors in the sense of system theory, i.e.states that are more likely to be entered than to be left. Such states tend to accumulate occupants over time (= diachronically).
     Around 70 languages from the above groups (except Yukagir for which only contemporary data are available) were analysed diachronically, to see how their personal pronouns changed over time.
     One conclusion is that `m' in 1sg increases over time, but 2sg `T' decreases over time. The increase in `m' was mainly due to the change of `b' in 1sg nom to `m' in the Mongolic and Tungusic languages (with a half-life time of 800-1000 years) and replacement of the `ǵ(h)' by `m' in the Indo-Uralic languages (with a half-life time of 3000-4500 years), both from `m' in the oblique cases. Another conclusion is that if a language has `m' in the 1sg it certainly has it in the 1sg obj; Kartvelian is the only exception. [DG: I think the Georgian mi/shen is an intruder, possibly copied from some early Turkic language.] Many other detailed conclusions are drawn.
     The author differentiates between steppe language spread (Eurasia), technology-driven language spread (e.g. Celtic) and imperial language spread (Rome, Greece, China). In the first type of spread, people are often multilingual and have little language loyalty, and in the last type of spread people are usually monolingual and have strong language loyalty. This may explain differences in replacement speeds in different languages.
     The author claims "Inheritance from a single ancient ancestor cannot explain the `m'." One reason is that the `m' in Altaic derives from *b, and therefore is not ancient. [DG: No, the `m' in 1sg nom derives from the `m' in 1sg obl, which was already there. This oblique `m' could well be a common ancient ancestor.]

* Whitman, J., The Relationship Between Japanese and Korean, 2012, pp. 18.
In reaction to Vovin's "Koreo-Japonica: A Re-evaluation of a Common Genetic Origin" (2010), which rejects almost all Korean-Japanese etymologies produced thus far, the author claims that by using improved phonetic rules and taking into account the root shape, it is possible to save a number of these etymologies. Some of the phonetic rules are predictive, as Vovin demanded, e.g. "If pK h is followed by i (or y), pJR has s, otherwise it has k." A total of 50 etymologies is produced, including nouns, pronouns, numerals, and body parts. [DG: A surprising etymology is, in Table 2.11, pJR mi(t) `three' ~ pK myech `a few'. Korean myech is a question word and means `how much', and means `a few' only because Korean question words can also be used as indeterminates (nugu `who', `somebody'). This makes it impossible that it relates to `three'.]
     There is also a heroic attempt to match Korean and Japanese numerals 1-10. [DG: There is really no need to: Turkish, Mongolian, and Tungusic, of similar time depth, don't have shared numerals either.]

* Ko Seongyeon, Tongue Root Harmony and Vowel Contrast in Northeast Asian Languages, (thesis), 2012, pp. 457.
An extensive and information-dense work.
In Contrastive Hierarchy Theory the contrastive features of vowels in given language ([low], [rounded], etc.) are ordered in a hierarchy. This places the vowels in a non-redundant tree structure, rather than in the usual table, which often contains redundant/useless entries; for an example see below. For CHT see Dresher `The Contrastive Hierarchy in Phonology' (2009), which does not give a method to obtain the hierarchy. The author orders the features by their phonetic activity. [DG: I don't get this. The most active/influential feature in (Modern) English is [voice]: voiceless consonants change the plural ending -s to -z, and the past tense ending -d to -t. But if your first split on the English consonants is along [+−voice], you get a weird tree!]
     Diachronic changes in vowel systems can often be attributed to changes in the feature hierarchy, or to losing a feature or acquiring a new feature.
     This thesis analyzes the vowel systems of the northeast Asian languages using Contrastive Hierarchy Theory, extended with the feature "Retracted Tongue Root" [RTR]. The analysis is performed on 12 Mongolian, 1 Korean, and 15 Tungusic languages. This allows a number of simple statements to be made that explain complicated questions.
     Tongue root position as a phonetic feature comes in two versions: advanced versus normal (+ATR/-ATR) (mainly in Africa), and normal versus retracted (-RTR/+RTR) (in northeast Asia).
     Traditionally Old Mongolian (13th century) is assumed to have palatal vowel harmony, which in modern times has shifted to pharyngeal (RTR), except in Kalmyk and Oirat. Based on the analyses done on the individual languages the author argues that it is more plausible that the RTR is original and Kalmyk and Oirat switched to palatal. A detailed description of the developments in the various languages is given.
     The vowels inventory of Old Mongolian is reanalyzed from an RTR point of view. This creates a good explanation for the way Mongol loan words were written in Hangul; makes the developments in the various daughter languages easier to understand; and creates a new classification of the Mongolic languages, making Kalmyk/Oirat a separate branch of Central Mongolic.
     The general contrastive hierarchy for Mongolic languages is [coronal] > [low] > [labial] > [RTR], but not all features are present for all languages. The Monguor languages differ from the other Mongolian languages in that they have lost the RTR feature. This reduces their number of vowels to 5 (from 7 to 14), although traces of RTR remain in the consonant (q vs. k). The Dagur languages, on the other hand, have lost the low/high contrast, also reducing their vowel set to the same 5 vowels, but their mapping from the Old Mongolian vowels is different. Kalmyk/Oirat replaced [RTR] by [dorsal].
     Middle Korean had 7 vowels, i, ɨ, u, ə, o, a, and ʌ, but the ʌ turned into ɨ, a, o or sporadically ə over time, leaving a 6-vowel system. The Hunminjeongeum (from (1446!) describes the seven vowels explicitly using the term "tongue retraction", and designates ʌ, o, and a as tongue-retracted; ɨ, u, and ə as slightly retracted; and i as not retracted.
     Vowel harmony in Middle Korean manifests itself in 1. all vowel in a word must have the same RTR state; 2. vowel-initial suffixes come in two variants, one for each RTR state. There are three possible pairs: u~o, ɨ~ʌ, and ǝ~a (-RTR~+RTR). [DG: This suggests that there were originally 3 vowels, + i]. Remarkably, some stems with only i take -RTR suffixes, while others take +RTR suffixes, and 6 stems occur with both types. It has been suggested that long ago there were two i-s, as in English `pill' (-RTR)and `peel' (+RTR) [DG: although this is an wrong comparison: in English the tongue body moves, not its root].
     The traditional feature set of the Korean vowels is [coronal], [high], [low], and [labial] (rounded). If we add the RTR feature, which is necessary to explain the harmonic behaviour of the suffixes, we have 5 binary features, which are good for 2^5 = 32 vowels, which is a bit much for 7 vowels. A contrastive hierarchy analysis shows that [coronal] > [low] > [labial] > [RTR] and that [high] is superfluous. The resulting tree (page 204) is

[+coronal] /i/
		[-RTR] /ǝ/
		[+RTR] /a/
			[-RTR] /ɨ/
			[+RTR] /ʌ/
			[-RTR] /u/
			[+RTR] /o/
[DG: This is the tree as it is shown in the book, including the reversed order of + and − in the lower (right-most) levels. I also think that a and ǝ are reversed and so are ʌ and ɨ; no sound in Korean has more tongue retraction than ɨ, and it gets [-RTR] here. Now it would be easy to dismiss this as a typographical glitch, but the assignments as above are confirmed in the text. I don't get it.]
     The change from Middle Korean to Early Modern Korean involved the disappearance of ʌ; the appearance of ɛ and e; and a serious reduction in vowel harmony. All three phenomena are examined and explained in great detail, sometimes supplying different scenarios for different dialects.
     The Korean Vowel Shift Controversy (pg. 229)
In the 1960s an attempt was made to determine the sounds of the Hangul letters in Middle Korean by examining how they were used to write Old Mongolian loan words, and comparing them to the known pronunciations of Old Mongolian. It was found that their vowels were shifted with respect to modern pronunciation, which led to the idea of the `Korean Vowel Shift', a 13 to 15 century rotation of the Korean vowel system, somewhat similar to the 14 to 16 century Great Vowel Shift in English. This idea was criticized right from the start:
1. It conflicted with the internal reconstruction of the Modern Korean vowels. This internal reconstruction was produced by comparing the vowel sets of seven Korean dialects and determining that no dialect preserved the pre-vowel-shift situation, which is unlikely.
[DG: No English dialect preserves the pre-vowel-shift situation either.]
2. It caused the description in the Hunminjeongeum to make no sense.
3. It could not explain vowel harmony in Middle Korean.
[DG: Yes, it can: the Early Middle Korean regular palatal vowel harmony system was warped by the Korean Vowel Shift (the author calls this explanation "undesirable").]
and four more similar objections follow.
     Next the author reconstructs the vowel set of Old (=Middle) Mongolian by combining the vowel sets of 13 modern Mongolian languages by "applying the `majority wins' principle" (pg. 237). An RTR analysis of this vowel set results in a diagram that is almost identical of that of Modern Mongolian (diagram 68). On the basis of this Middle Mongolian RTR vowel set and the Middle Korean RTR vowel set, the author gives reasonable explanations of the observed mapping from Middle Mongolian vowel onto Middle Korean vowels without needing a Korean Vowel Shift.
     Since the end of the 1990s the Manchu vowel system has been analyzed using advanced tongue root position (ATR) as a feature. The author shows, however, that a description using RTR is more appropriate, and is a viable option for all Tungusic languages.
     The contrastive hierarchy for Tungusic languages is almost invariantly [low] > [coronal] > [RTR] > [labial], except that Oroqen uses [low] > [coronal] > [labial] > [RTR]. This differs from the contrastive hierarchy of Khalkha (Standard) Mongolian only in the order of the [low] and [coronal] features. The author derives from this difference the differences in labial component of the vowel harmony in both languages (after rejecting earlier proposals).
1. The contrast-driven typology of the Altaic languages helps to explain the differences in vowel harmony between the languages.
2. Although Turkic vowel harmony (not discussed in this thesis) is based on the [palatal] feature, Mongolic, Tungusic and Korean have vowel contrasts based on [RTR], so it is likely that proto-Altaic had the same, and that proto-Turkic has switched to [palatal], perhaps along the same way Kalmyk and Oirat have.
3. The contrast-driven classification of the Altaic languages seems to correctly reflect the genetic/geographical affinity between languages.
4. The fact that Mongolic, Tungusic, and Korean use 4 features and Turkic 3 may underlie the difference between the asymmetric vowel sets of the first three languages and the symmetric set of the last.

* Kang Hijo, Ko Seongyeon, In Search of the Acoustic Correlates of Tongue Root Contrast in Three Altaic Languages: Western Buriat, Tsongol Buriat, and Ewen, 2012, pp. 25.
Voice spectra were recorded for tongue root sensitive words in Western Buriat and Tsongol Buriat (both Mongolic), and Ewen (Tungusic). It was found that formant F1 was lower for [+ATR]/[−RTR] than for [−ATR]/[+RTR]. Formants B1 and A1−A2 were also affected, but that may have been caused by the change in F1. It is not clear that there were no other, not registered, effects.

* Sagart, L., How Many Independent Rice Vocabularies in Asia?, 2011, pp. 20.
Almost all East-Asian languages have reconstructible rice-vocabularies; the exceptions are the micro-Altaic languages.
     Korea and Japan obtained their rice culture after they separated, if they ever were related. Japan obtained it from a people speaking a pre-Austronesian language from the Sino-Tibetan-Austronesian macro-family, which supplied the word kome. (The author has established the Sino-Tibetan-Austronesian macro-family on the basis of sound correspondences between the main syllable in Old Chinese words and the last syllable of Austronesian words and tentatively equates proto-Sino-Tibetan-Austronesian with Middle Yangshao, c. 4500 BZ. It can also explain the sharing of japonica rice by the Sino-Tibetans and Austronesians.)
     After comparing dozens of word list and considering dozens of proposed theories the author concludes that there (at least) two rice culture vocabularies: one proto-Austroasiatic; and the rest of the rice vocabularies of Asian languages, based on Sino-Tibetan-Austronesian. The rice vocabulary of India does not show any relation with either of these.
[DG: Perhaps that is why the author writes "at least two".]

* Vovin, A., Why Japonic Is Not Demonstrably Related to `Altaic' or Korean, 2011, pp. 8.
Frontal assault on Robbeets ("Is Japanese related to Korean,Tungusic,Mongolic and Turkic?", 2005) and Unger ("The Role of Contact in the Origins of the Japanese and Korean Languages", 2009).
     The main objections to Robbeets are:
1. Regular productive-predictive correspondences have never been demonstrated. [DG: I suppose this means that the correspondences are not of PIE quality. Well, we have IE material from 3600 years ago, and the oldest Turkish text is 1300 years old and the rest is even more recent. There is only one Indo-European.]
2. Only fragments of word are compared, f.e. only the first CVC sequence.
3. There is no common paradigmatic morphology.
     The "Japanese is Altaic" hypothesis is presently defended most forcefully by Robbeets. The author examines one denominal verbal suffix, -la-, which Robbeets claims exists in proto-Japanese, Old Japanese, Turkic, Mongolic, and Tungusic. Pointing to the above faults, the author rejects Robbeets' 5 examples for Old Japanese; the Turkic examples are acceptable, but the Mongolic examples are probably loans from Turkic; and many Manchu examples are loans from Mongolian that ended in l and obtained an additional -a in the process.
     The "Japanese is related to Korean" hypothesis is presently best represented by Unger. So the author presents 37 Korean/Japanese correspondences from Unger, and demonstrates that there are no consistent sound correspondences between them, by showing a list of unaccounted elements, comparisons to ghost word, and incorrect or ‘polished’ semantics found in these correspondences.
     Finally the author shows that 6 words are sufficient to show that Malay and Hawaiian or (Tokyo) Japanese and Shuri (Ryukyuan) are related, and invites the audience to find 6 such words for Korean and Japanese. [DG: Is he serious? The time depth of M/H or J/S is at most 2000 years; that of K/J is more than 5000, if they are related at all.]
     The paper also contains a list of the "primary sources" consulted by the author:
For Japanese: Kojiki kayō, 712 AD; Komonjo 25, 762 AD; Man’yōshū, ca. 759 AD; Nihonshoki kayō, 720 AD.
For Korean: Kwukuppang enhay, 1466 AD; Yongpi echenka, 1445 AD.
     Needless to say, the "secondary sources" are publications by contemporaries.
     There is also a set of 32 slides of a talk given under the same title by the author on July 30, 2011, in Osaka.

* Vovin, A., First and Second Person Singular Pronouns: A Pillar or a Pillory of the ‘Altaic’ Hypothesis?, 2011, pp. 28.
The Old Turkic, Old Mongolic and Old Tungusic (Inner 'Altaic') 1st and 2nd pronouns singular can be summarized thus:

			Old Turkic	Old Mongolic	Old Tungusic
	1ps nom.	bän 		bi		bi
	1ps obl.	min-		min-		min-
	2ps nom.	sän		ti		si
	2ps obl.	sin-		tin-		sin-
The same table for the West-Germanic languages, with a time depth of about 1500 years, for comparison:
			English		Dutch		German
	1ps nom.	a:j		ik		iç
	1ps obl.	mi:		mεj		mir/miç
	2ps nom.	ju:		jεj		du
	2ps obl.	ju:		jɔw		dir/diç
This is such strong evidence for a genetic relationship between Old Turkic, Old Mongolian, and Old Tungusic, that it requires an explanation from anyone who wants to deny such relationship. And that is exactly what this paper tries to do.
     The author observes that Old Mongolian and Old Tungusic 1ps and 2ps are very similar, much more similar than the rest of Old Mongolian and Old Tungusic, which suggest they are far newer than the rest of the languages. He then argues that these pronouns are close to those of Old Bulghar Turkish (without giving the Bulghar pronouns). The conclusion is that the Old Mongolian and Old Tungusic pronouns are copies of those of Bulghar Turkish, of recent date (~ 2000 years ago). [DG: In view of the changes in the West-Germanic pronouns in about 1500 years, I'd update that to at most a 1000 years, but then Bulghar Turkish and Tungusic were no longer in contact. In short, they are even too new for that.]
     [DG: Copying pronouns is possible but very rare. I find it hard to believe that two languages should have done so simultaneously.]

* Ko Seongyeon, Vowel Contrast and Vowel Harmony Shift in the Mongolic Languages, 2011, pp. 22.
Traditionally Old Mongolian (13th century) is assumed to have palatal vowel harmony, which in modern times has shifted to pharyngeal (RTR), except in Kalmyk and Oirat. Based on the analyses done on the individual languages the author argues that it is more plausible that the RTR is original and Kalmyk and Oirat switched to palatal. A detailed description of the developments in the various languages is given.

* Ko Seongyeon, The End of the Korean Vowel Shift Controversy, (talk), 2011, pp. 47 slides.
Very informative paper, whether one agrees or not.
Middle Korean had very clear vowel harmony, and the 16th century Korean grammarians had no trouble explaining it. Middle Korean had 7 vowels: 3 Yang ones, /a/, /ʌ/, /o/, 3 Yin ones, /ə/, /ɨ/, /u/, and one neutral, /i/. The Yang and Yin vowels came in pairs:

    /a/ ㅏ ~ /ə/ ㅓ
    /ʌ/ ㆍ ~ /ɨ/ ㅡ
    /o/ ㅗ ~ /u/ ㅜ
Even the Hangul letters showed their agreement. Words contain either Yang vowels or Yin vowels. A number of endings come in two varieties, one for Yang words an one for Yin words: the stem extender -a/, the adnominal suffix -on/-un, the accusative particle -ʌl/-ɨl, and the locative particle -aj/-ǝj.
     With the advent of phonetics in the West, the question arose how to characterize the difference between Yin and Yang in phonetic terms. Vowels are normally ordered in a triangle, with features like [high], [front], etc. But a characterization of Yin and Yang in terms of such features appeared impossible. Also, when investigating how Mongolian loan words were transcribed into Korean and how Korean words were transcribed in Chinese, many discrepancies were found. One peculiarity was that the Middle Mongolian u was transcribed with an ㅗ (o). Research into this subject resulted in a vowel assignment for the Hangul characters that was different from the present one:
    ㅏ = a, ㅓ = e; ㅗ = u, ㅜ = ü; ㆍ = ɔ, ㅡ = ǝ; ㅣ = i
And lo and behold, this arrangement showed clear palatal (umlaut) vowel harmony: e is a umlaut, ü is u umlaut, ǝ is ɔ umlaut.
     Now a new question arose: how did this Middle Korean arrangement evolve into the present one? The process is called the "Korean Vowel Shift", and the present paper gives an excellent pictorial explanation of it, based on Lee (1977). The Great Korean Vowel Shift (there is also a Small Korean Vowel Shift, from Old Korean to Middle Korean) has five out of the seven vowel doing musical chairs: e → ǝ → ü → u → ɔ, with only a and i remaining wallflowers.
     There are, however, many problems with this fortuitous solution; some are:
• The descriptions of the sounds given in the `Hunminjeongeum Haerye' no longer make sense.
• No dialect preserves any trace of it, which is hard to believe.
• All steps are phonetically unmotivated.
• The steps are ridiculously complex: fronting, backing, raising, lowering, unrounding, etc.
• The shapes of the Hangul letters no longer make sense.
     Also, some of the evidence for the Korean Vowel Shift is weak:
• The data from Middle Mongolian are inconsistent: Middle Mongolian olang, bora, ɣodoli should be transcribed using ɔ ㆍ but are instead attested as 오랑, 보라, 고도리.
• The data from Middle Mongolian is unreliable: even in the 18th century the compilers of the textbook `Monge Nokeltay' still used Korean u ㅜ to transcribe Mongolian ü, although the Korean Vowel Shift was long completed.
• Even today Mongolian u is often heard as ㅗ o by Koreans.
• The Chinese data are unreliable, because Chinese underwent a vowel shift too (ou, uü) around the same period.
But if no Korean Vowel Shift, then what?
     In the next 14 slides the vowel harmony in Old Mongolian is examined. It is found to be based on the feature "Retracted Tongue Root" [+−RTR]. Since it has been determined that a, ʌ, o have [+RTR], and ə, ɨ, u have [−RTR], the Yang-Yin difference can be characterized as the RTR feature, and it may be concluded that there was no Korean Vowel Shift. [DG: How was that determined? It certainly does not correspond to my experience. There is no vowel that requires more tongue root retraction than ɨ, and yet it is marked as [−RTR]. Very unsatisfactory.]

* Antonov, A., Turkic `kümüš' `silver' and the Lambdacism vs Sigmatism Debate, 2011, pp. 28.
Casting the net wide, the author finds cognates for Turkic kümüš ~ Chuvash kӗmӗl `silver' in tens of Sino-Tibetan and Austro-Asiatic languages, thereby proving that it is a "Wanderwort". In all of these languages the word ends in l or ɬ (the suggested value for ), except in Common Turkic, where it has an š. In none of the languages an etymology for the word was found, so the origin of the word is not known.
     The author is careful not to draw a stronger conclusion than that attributing the proto-Turkic ~ Mongolian l correspondence to loans is not the only possibility.
     [DG: That seems to weak to me:
1. If the word originated outside proto-Turkic, it ended in an l-sound, entered into proto-Turkic with that l-sound, and Common Turkic obtained its š through sigmatism.
2. If the word originated inside proto-Turkic, it may have ended in š, and Chuvash may have obtained its l through lambdacism. Then the Wanderwort must have spread from Chuvash. But this is highly unlikely, because all cognates in Sino-Tibetan and Austro-Asiatic have an u-sound in the last syllable (a fact not pointed out by the author), and these u-sounds can hardly have originated from Chuvash ӗ. Also, the spilt between Common Turkic and Chuvash was probably too late for the word to spread this widely.
3. If the word originated inside proto-Turkic and ended in l, we are back at situation 1 above.
4. Since it it highly unlikely that 2 holds, it is highly likely that either 1 or 3 holds. In both cases there was sigmatism between proto-Turkic and Common Turkic for this word. ]

* Róna-Tas, A., On Rhotacism, 2011, pp. 5.
Odd notes on rhotacism.
     There are three types of rhotacism, nr (Northern Albanian zâni → Standard Albanian zëri `voice'), s/zr (Latin musmuris `mouse', Dutch ijzer → English iron) and t/dr (Old Turkish *adak → *aδaγura `foot', and in `pottage' → `porridge'). A fourth type occurs in Dahur, where all consonants which close a syllable became r: Mongolian nabči → Dahur narči `leaf'.
     All proposals for rhotacism in the Altaic languages start from an r-like sound that became z in Common Turkic and r in Chuvash and the other Altaic languages. The author uses ♣ `club' and ♠ `spade' for for , to no added value.
     The author gives tens of examples of rhotacism in Old Turkic, e.g. East Old Turkic ikiz `twin' → Turkish ikirčgü `doubt'. [DG: Other people say that there was zetacism here, and the r in the second word was protected from zetacism by the following consonant.]
     The z in biz `we' is often interpreted as a dualis marker. If so, it is actually a marker for the singular of objects that occur in pairs: köz `eye'. Many such words are body parts but there are also non-paired body parts that end in -z: bogaz `throat'. So it is better to conclude that there are two markers z, one for body parts and one for forming a plural.
     Because the word for `stirrup' had z in some Turkish languages (izeŋü) and r in Chuvash (ireŋü), the split between Common Turkic and Chuvash must have occurred after the stirrup was invented (3rd century BZ for the leather stirrup, 3rd to 5th century AZ for the iron one).
     The author lists 16 Hungarian loans from Common Turkic (some only from Tatar) that have r where Common Turkic has z; these were borrowed before the onset of rhotacism. [DG: I still think this should be "zetacism'. Am I misreading something?] Nine similar loan words with z follow; the earliest attested is from the year 1111. So it seems that everything we are talking about here happened after the year 0.

* Janhunen, J., Reconstructing the Language Map of Prehistorical Northeast Asia, 2010, pp. 23.
This paper is the reflection of all the author's knowledge of Northeast Asian languages, connecting all the dots. In line with this, the paper has no list of referenced works. This is at times annoying because it leaves many claims hanging in the air. For example: "There are also native cultural words in Ghilyak that have no parallels in Tungusic or other languages." (pg 294). Is that info from published research? From personal observation? The paper reads as the science supplement of a prominent newspaper: fascinating information, no hard data.
     The languages of Northeast Asia come in 8 phyla (lineages):
1. Sinitic. The relative uniformity of Mandarin suggests that it is a recent expansion from a limited area. Useful written material in Chinese starts in the 7th century.
2. Turkic, consisting of Bulghar Turkic (1 language) and Common Turkic (the rest). Again the uniformity of the Common Turkic languages suggests a recent break-up. Oldest available arterial dates from the 6th century.
3. Mongolian, consisting of the mutually more or less intelligible Common Mongolian languages + a number of outliers. Oldest material dates from the 12th century.
4. Tungusic, over a very large chunk of Siberia. Again 12th century (Jurchen).
5. Amuric, better known as Nivkh or Gilyak or Ghilyak, with three dialects along the Amur river and on Sakhalin. No data on older forms.
6. Koreanic, very uniform inside Korea, + a dialect/language on Jeju. The oldest material is from the 12th century.
7. Japanese, consisting of Japanese proper (1 language) + Ryukyuan with half a dozen closely related languages. The oldest material is from the 6th century.
8. Ainuic, the collection of Ainu dialects. There are no data on older forms.
     Using internal reconstruction and the comparative method proto-languages of all 8 phyla can be constructed. This pushes back the age of each proto-language by 500-1000 years. All of these phyla are shallower than the families of PIE, and the author argues that PIE and the other great phyla (Semitic, Austronesian [DG: + Bantu] were vigorously expanding people, and there were none such (except the Sinitic) in Northeast Asia. [DG: Another important point seems to me that the PIE people and the Semitic people began to write early.]
     The fairly even time depth of all 8 phyla allows us to place the home lands of the phyla on the map of Northeast Asia of 1500 to 1000 years ago; the definition of `home land' in this context is the location of the proto-language, i.e. the oldest reconstructible form of the language. There is a rule of thumb saying that the geographical origin of a phenomenon is at the place of its greatest variety, but this rule has to be used with care: the greatest variety of Turkic languages can be found in the steppe, but Turkic loan words in Mongolian indicate that the Turkic home land was much farther to the east. The home land of Sinitic was the Yellow River Basin, from which the Sinitic languages radiated in all directions; some of those that went south became the Tibetan (Bodic) languages, one of those that went north reached Manchuria, from where it spread over North China and became Mandarin. Loan word evidence suggests that the Mongolian home land was in southern Mongolia, between the Turkic home land on the west and the Tungusic home land on the east. Tentatively, with many caveats, Turkic could be associated with the Xiongnu, Mongolic with the Donghu, and Tungusic with the Sushen. The author places the home lands of Koreanic and Japonic in the south of the Korean peninsula (see the definition of `home land' above). The Amuric home land lay probably in Manchuria, north of that of Tungusic, and Ainu was one of the probably hundreds of languages in Jomon Japan.
     The above map was constructed mainly on the basic of loan words, i.e. lexical evidence. It brings together the Turkic, Mongolic, Tungusic, Amuric, and in a lesser sense Koreanic and Japonic together on a compact area, where they have no doubt influenced each other. This resulted in the "Altaic structural type", exhibited by 5 of the 6 phyla. Only Amuric is not emphatically Altaic; the author suggests that Amuric was influenced by some now extinct languages of a non-Altaic and non-Sinitic type structure, and that the Altaic features in Amuric are recent.

* Robbeets, M., Transeurasian: Can Verbal Morphology End the Controversy?, 2010, pp. 34.
The (macro-)Altaic languages are suddenly called "Transeurasian languages". without introduction or argumentation.
[DG: To take a well-known technical term and to arbitrarily replace it with one of one's own is unheard of. No argumentation is given. Taking the first meaning of "Trans" as `beyond' the term would mean `the languages that lie beyond Eurasia', the first of which would be Hawaiian. Taking its second meaning `from one end to the other' (as in "Transsiberian Railroad") the meaning would just be `Eurasian Languages', and include Dutch, Basque, Nivkh, and many others. In short, it is a misnomer, and probably intended to appropriate the subject.]
     Both Vovin (2005) and Dybo & Starostin (2008) claim to have ended the Altaic controversy, unfortunately in diametrically opposite ways. But they agree on one thing, evidence of common paradigmata would be best.
     The author argues that it is unreasonable to expect paradignatic evidence from the Transeurasian languages because of the great time depth, more than a millennium more that for PIE, and because the real proof of PIE rests on the classic languages, with paradigms known from some 2500 years back, whereas the oldest data from the Transeurasian languages is at best 1500 years old. But the author also points out that even the modern PIE languages hold enough paradigmatic information to recognize PIE, citing the paradigmata of Spanish, German, and Iranian (Modern Persian). [DG: I don't understand this argument. If anything it shows that we should expect paradigmatic evidence from the Transeurasian languages:

		Spanish	German	Iranian	Finnish	Hungarian
	1sg.	-o 	-e	-æm	-n	-k/-m
	2sg.	-s 	-st	-i	-t	-l/-d
	3sg.	-ø 	-t	-æd	--	-ø/-i
	1pl.	-mos 	-en	-im	-mme	-ünk/-ük
	2pl.	-is 	-t	-id	-tte	-tek/-itek
	3pl.	-n 	-en	-ænd	-vat	-nek/-ik
The first three columns shows that paradigmatic relatedness still shows after 5000 years. If we add columns for Finnish and Hungarian, we see paradigmatic evidence for Indo-Uralic, at least another 1000 years older. Discard the plural marker -k from Hungarian, and possibly the plural marker -s from Spanish, and you get an even better fit.]
     Therefore the author turns to other morphological evidence: actionality suffixes. They occur in all Transeurasian languages, and the author has reported on them in "How the Actional Suffix Chain Connects Japanese to Altaic" (2007). These actionality suffixes are especially unlikely to be borrowed. and have high evidential power. In summary this common verbal morphology belongs to the "very thin layer" that holds the Transeurasian languages together as a genetic unit.

* Ko Seongyeon, Vowel Contrast and Vowel Harmony Shift in the Mongolic Languages, (talk/presentation), 2010, pp. 17.
Point-by-point summary of the eponomous paper of 2011.

* Vovin, A., Koreo-Japonica: A Re-evaluation of a Common Genetic Origin, 2010, pp. 298.
Comparisons between Korean and Japanese have been based on reconstructed forms made 30 years ago or longer, and reconstructions based on the most modern theories should be used. An example from Korean: there regular and irregular p-verbs. Traditionally the irregular verbs are explained by lenition, but a more likely view is that all medial single p were lenited in principle, but some were preceded by and protected by a consonant. This makes the regular p verbs originally disyllablic, which changes their properties in comparisons to Japanese. Interesting material is presented to support this and similar, ideas, also by comparing related effects in Nivkh. For Japanese, the vowel system is revised, for Ryukyuan some consonants.
     In Chapter 2 proposed shared morphological items are examined, and the author finds none that pass muster. Remarkably, the obvious shared morpheme, -ka/-ga marking subjects in both languages, is not mentioned.
     Chapter 3 examines the 347 etymologies proposed by Whitman (1985). The conclusions are: 11 possible cognates, 75 obvious loans (from Old Korean to Central Japanese), and 261 rejected etymologies. The list of the 11 potential Korea-Japonic cognates is then presented; it can only be described as a sorry lot, with no body parts, and only one basic word: `fire'.
     The conclusion is that any lexical similarity between Korean and Japanese most likely results from loans. The 11 possible cognates can be real cognates (in which case Korean and Japanese are genetically relates) or can be very old loans.

* Vovin, A., Is Japanese Katakana Derived from Korean `Kwukyel'?, in Contemporary Korean Linguistics--International Perspectives, 2010, pp. 7.
The author demonstrates that katakana and the Korean syllabary kwukyel (gu.gyeol, 구결) are related by showing a list of 20 exactly matching pairs. A second list contains 21 matching pairs between man'yogana and kwukyel. Although the oldest known occurrences of both katakana and man'yogana are some centuries before those of kwukyel, the author claims that the Japanese scripts are based on the Korean kwukyel because "kwukyel is more polyphonic than katakana", which is apparently intended to mean that some kwukyel signs have more than one phonetic value. [DG: I'm not sure I follow. Perhaps the "polyphony" shows that kwukyel was not yet fully stabilized and therefore older?]

* Vovin, A., Once Again on the Ruan-ruan Language, 2010, pp. 10.
Referring to Vovin `Once Again on the Etymology of the Title qaγan' (2007), the author clarifies qa-n as deriving from γa from proto-Yeniseian *qʌ:j or *χʌ:j `ruler', and -n from Mongolian for singular; and the female forms marked by the suffix -tu- are a Ruan-ruan innovation. [DG: This is adding complication to complication. The simplest explanation of qaγan is reduplication of a noun qa `ruler', much like "Shahanshah". The name of the language already shows that it is not averse to reduplication.]
     To find more loans in Old Turkish that could be from the Ruan-ruan language the author examines collocations (combinations like "each and every") in Old Turkish where one member is clearly of Turkish descent and the other is not recognizable from any surrounding language. [DG: Why these would be better than just unidentifiable Old Turkish words eludes me.] This yielded 4 words possibly attributable to Ruan-ruan. In the same vein 2 words are found that are strange an unusual synonyms for common words.
     There are separate words for the tens from 20 to 50; 60 and 70 are expressed by the suffix mIš; and 80 and 90 as 8 on and 9 on. The word on `ten' is known everywhere in the Turkish languages, but there is no other known use of mIš for `ten' anywhere. Again the word is ascribed to Ruan-ruan.
     Irregular vowel sound changes in loans from Middle Chinese become regular when one assumes that some words were borrowed directly from Middle Chinese and some others through intermediation of an unknown language, which is an indication that there was an unknown language around from (or through) which Old Turkey borrowed words. Again Ruan-ruan is suggested.

* Alonso de la Fuente, J.A., Urban Legends: Turkish kayık `boat' and “Eskimo” qayaq ‘kayak’, 2010, pp. 18.
The author shows that Turkish kayık `boat' comes ultimately from proto-Turkic *kad- `fir tree', and Greenlandic qan-yaq comes ultimately from proto-Eskimo qan(ə)- `to go/come'. Consequently the two words have no relationship.

* Greenhill, S.J., The Shape and Tempo of Language Evolution, 2010, pp. 9.
Language evolution, in particular of typological features. The paper consists of two parts, the first concerned with the question how predictive of language relationships typological features are, and the second with the question how stable typological features are diachronically.
     In the first part 136 typological features of 99 well-attested languages are correlated and presented in a NeighbourNet representation. This results in a figure with two well-defined lobes and a fuzzy fan, separated from each other by clear gaps. Lobe 1 contains almost exclusively languages from Northern Eurasia; lobe 2 contains primarily languages from the rest of Eurasia and from Africa; and the fuzzy fan more ore less covers the rest of the world: the Americas and Australia. The classification is far from perfect: Quechua is allocated in lobe 1, next to Hungarian, and although the micro-Altaic languages are grouped together in lobe 1, Korean and Japanese are in the fan part, between Burmese, Nama (!), Nivkh, and Ainu. But the classification is not random either, which, according to the authors, may point at deep underlying relationships.
     In the second part two well-analysed phyla are considered, the Indo-European and the Austronesian phylum. Estimates of the dates of the nodes in the family trees of these phyla are known. These dates are an indication of how well features are retained. By computing the retention index for a given feature, it can be determined how well that feature is retained. The retention index is an indication of how similar a tree based on the feature alone is to the known tree, including the dates on the nodes.
     The retention indexes for about a hundred words from the lexicon and for about a hundred typological features were computed for the Indo-European and the Austronesian phylum, and compared. The result was that for Indo-European the lexicon and the typology were about equally good predictors of the tree structure, and for Austronesian typology did somewhat worse than the lexicon.
     The conclusion is that typological features are not generally retained significantly longer than lexical features (words), as is sometimes claimed.

* Turchin, P., Peiros, I., Gell-Mann, M., Analyzing Genetic Connections between Languages by Matching Consonant Classes, 2010, pp. 10.
53 Eurasian and North African languages are considered, and for each of these the Swadesh 100 words list is obtained. Consonants are mapped onto 9 groups as per Dolgopolsky. Each group is represented by a symbol, and words are reduced to the symbols of their first 2 mapped consonants. This reduces all words to two-symmbol words, and there are only 81 such words in the world.
     To determine the relatedness between two languages we count the number of words in their 100-word lists that have the same (two-symbol) form and the same meaning. This number is the similarity percentage (Similarity Index, SI). The question is, however, at which percentage this similarity becomes statistically significant. If the languages are unrelated, the word lists would be random with respect to each other and on would expect to find 100/81 matching words on the average [DG: is this correct??]. But even if the list are unrelated, the words are not independent: consonant frequencies and consonant cluster frequencies in the languages will be non-uniform, raising the number of matching words.
     To compensate for this, the interdependency between the consonants is determined as follows. A 100 times a random word from one list is compared to a random word in the other list, and the matches are counted. (This is slightly more complicated than comparing the words in the two lists in random order, because it can handle incomplete lists) . This procedure is repeated 10,000 times. This yields the frequency distribution of the Similarity Index for independent languages with the given consonant dependencies. [DG: The authors call this procedure "bootstrap method (15)" for reasons that are not clear to me.] Next the two lists are compared in the correct semantic order, and the number of matching words is noted. The area to the right of this number in the probability distribution is the chance that the detected similarity is due to chance. When that probability turned out to be 0, the procedure was repeated 1,00,000 times. Even then for known related languages the probability remained 0.
     Results: Languages from the same family have an SI of between 45 (English-Icelandic) and 76 (Dutch/German), and P < 10^−6; languages from the same phylum but a different family have an SI of between 1 (Nepali-Icelandic) and 7 (e.g. English/Bengali), and 0.005 < P < 0.7. The Semitic languages (5 of them) all get P < 10^−6, and so do proto-Iranian, proto-Slavic, proto-Baltic, and proto-Germanic [DG: not a varied choice.] The four Altaic proto-languages (Korean is not considered) get 10^−5 < P < 0.002 (the latter, the worst match, for proto-Turkic/proto-Tungusic).
     This shows that the four Altaic families are more alike than chance allows (but less than the Indo-European families), but the question remains whether this is from genetic or areal influence. An areal effect would show up as increased similarity between, for example, proto-Mongolic and Old Chinese (but we find P = 0.21), or between Turkish and Kurdish (but we find P = 0.10).

* Robbeets, M., Insubordination in Altaic, 2009, pp. 19.
Desubordination (a.k.a. insubordination) is a useful tool for observing genealogical continuity, because few languages have it, which makes it specific. (But see Malchukov `Verbalization and insubordination in Siberian languages' (2013).)
     The author identifies the verb endings -n, -rV, and -m in all 5 language groups of macro-Altaic, and supports this with extensive examples and analyses. The meanings of these endings are similar in all groups: -n -- factual / non-future tense; -rV -- supposed / future tense; and -m -- deverbal noun (as evident in Korean). These verb endings normally create non-finite verb forms (participles, converbs, etc.), which, however, in many of the languages involved, are also used as finite verb forms, i.e., show desubordination. The author argues that the parallel occurrence of desubordination of these endings in all five Altaic language group is a strong indication of their genetic relationship.
     [DG: I agree that these endings form one of the strongest indications that the macro-Altaic languages are genetically related in the end, but I am worried about the time scale. These endings are old in the proto-languages of their specific language groups, and probably already existed for say a 1000 years before that, which places them somewhere in the first millennium BZ. But proto-Altaic is placed around 5500 BZ, which leaves a gap of 5000 years in which the endings should have remained unchanged. Hard to believe.]
     In the last sentence of the paper the author manages to use the word `insubordination' in both meanings.

* Janhunen, J., Proto-Uralic: What, Where, and When?, 2009, pp. 22.
The author observes that the pre-historic state of languages is that which we find in New Guinea, Siberia before Russian rule, and other places on earth: many small languages, each with a few tens to a few thousands of speakers. Language phyla only came into being in and after the Neolithic Revolution, when an arbitrary language collected enough speakers to start a language phylum. The spread of the culture caused languages to be replaced, decreasing the number of languages; on the other hand, the replacing languages, covering larger areas now, differentiated, increasing the number of related languages.
     The author argues that the numerous small different languages of the Mesolithic era covered a very large area of uniform Mesolithic culture, and that hence culture cannot be reliably related to language.
     The author reconstructs the history of the Uralic languages based on linguistic evidence, rejecting perhaps a dozen previous hypotheses in the process. The author sees Proto-Uralic as a language with a Mesolithic vocabulary in Northern Asia on the edge of the steppe and the tundra, with a number of speakers in three or four digits. The first to split off were the Samoyeds, who went east. In the Neolithic, the remaining Proto-Finn-Ugric speaking people started moving west along the edge of the steppe and the tundra, leaving behind successively speakers of Mansi, Khanti, Permic, Mari, Saami, Mordvin and finally Finnish. The Mansi speakers moved south into the steppe, where the speakers of Hungarian split off and moved west. All this is cached in careful phrases, frequently using "the evidence suggests"; no explicit language examples given.

* Maher, J.C., A Brief History of Pidgins and Creoles in Japan, 2009, pp. 13.
The mainstay of this paper is the description of five creoles in Japan, but there is a good section on Japanese as a creole, elaborating on Maher (1998).
     Pre-Japanese has the following properties in common with creole languages:
1. restricted noun inflections;
2. conjunctions without conjunction particles;
3. development towards more surface differentiation;
4. subordinate clauses through nominalization;
5. a tense-aspect system as described for creoles (Bickerton 1981);
6. the use of reduplication for repetition and plurals.
     Pre-Japanese as a creole is explained by the following scenario: The Proto-Altaic Yayoi, entering North Kyushu from Korea, found there speakers of various Paleo-Siberian (from Sakhalin) and Austronesian (from Taiwan) languages. The great difference in technical level between the Yayoi and the Kyushuans necessitated a trade language understandable and learnable by all: a pidgin. Pidgins become the languages of the next generation, so soon a creole developed. [DG: This means that Japanese would literally be no more than perhaps 2000 years old and has a rather unique origin. This will please some and irk others.]
     [DG: It seems obvious that under this scenario Japanese cannot be related to Koguryo, because it appeared too late and in the wrong place; but we know it is related (Beckwith, 2004). However, the known relation between Koguryo is based in lexical items only, and a percentage will have survived later creolization. This means that Koguryo is related to Yayoi, rather than to Japanese.]

* Vovin, A., Japanese, Korean, and Other 'Non Altaic' Languages, 2009, pp. 42.
The author starts by quoting D. Sinor (1963): "If a scholar of Poppe's stature and knowledge fails to prove the theory of the genetic relationship of the Altaic languages, there must me something very wrong with that theory." [DG: We know more now; and we have computers.]
     The paper is a very critical analysis of Robbeets' thesis (2005). In her thesis Robbeets claims to sift through existing purported cognate etymologies, but the procedure is fundamentally flawed: it is performed from an a priori conviction that a relationship exists. This leads to the following specific defects:
0. Etymologies are borrowed uncritically from the EDAL.
1. Sometimes inflectional OJ morphemes are treated as derivational.
2. Some etymologies only work through incorrect segmentations.
3. The etymologies are not screened against cultural and historical context.
4. Etymologies are mostly borrowed from secondary sources, and do not go back to original text; and Old Turkish is ignored at all.
5. Robbeets is satisfied with matches for the initial consonant, the medial vowel and the medial consonant; for Japanese this does not cover the whole word.
6. No regular productive-predictive correspondences are given.
7. A fair amount of the etymologies taken from the EDAL and other sources contain words that do not exist (`ghosts').
8. Robbeets seems to consider borrowing only for nouns, not for basic vocabulary, verbs, or morphemes.
9. Some etymologies are semantically loose, e.g. `fire' ~ `to dry something in the sun'.
10. The book is full of anachronisms, incorrect citations, faulty transcriptions, unsubstantiated reconstructions, etc.
     The paper closes with 31 pages of detailed examples of these faults and blemishes.

* Georg, S., Review of Robbeets `Is Japanese Related to Korean, Tungusic, Mongolic and Turkic?', 2009, pp. 33.
Devastating critique, cached in scathing terms, sometimes confusing to read due to its liberal use of sarcasm and German-style industrial-strength sentences. Some points:
• The second part of the book, the “Etymological Index of Japanese”, has been constructed by copy-and-paste from existing literature, leaving the snippets completely without context.
• Criticisms to Altaic etymologies and the Altaic hypothesis in the literatur (e.g. Doerfer) are dismissed off-hand.
• Many of the rejected etymologies are rejected on semantic grounds, but the semantic check is according to Robbeets’ personal “semantic whim”.
• Although the Robbeets' etymologies consist of Japanese ~ Korean ~ Tungusic ~ Mongolic ~ Turkic words, only the Japanese and Korean data have been checked; the rest was copied as is. [DG: I'd suppose one should be able to rely on the work of others.]
• Only the first CVC of a word is considered, the rest is "discarded – or, better, swept under the rug". [DG: Robbeets explains this as part of the methodology.]
• The author pokes holes in some "core evidence" etymologies, e.g. the one for `tooth', where palV is presented as the Tungusic equivalent. But this word occurs only in two dialects, where it means `molar' and is a loan from Mongolian paluqa `hammer'.
     Robbeets' thesis ended with a list of 8 achievement; this critique ends with a list contradicting all 8 of them.

* Dresher, B.E., The Contrastive Hierarchy in Phonology, 2009, pp. 23.
[DG: This is relevant for the discussion of the vowel harmony systems of the Altaic languages.]
     In theory the phonemes of a language are distinguished from each other by features, which can each have two or more partially ordered values. In practice these features are not orthogonal to each other, so a matrix representation of them has a certain redundancy, and a search for a better, structured representation is in order.
     A naive approach starts from any pair of phonemes that differ in one feature only. That feature is certainly contrastive. and the set of phonemes can be split in N sets, where N is the arity of the feature. This process is repeated until each phoneme is in a set of itself (= all the sets are singletons). This arranges the phonemes in a tree, without redundancy.
     This "Pairwise Algorithm" fails when there are no more phonemes that differ in one feature only and not all sets are singletons. One can then continue by just grabbing a not yet used feature, split the sets along that feature, and continue in this way until all sets are singletons. Seeing that this algorithm works even when there are no pairs that differ in one feature only, we might as well ignore the pairs at all and start splitting right away. This yields the "Successive Division Algorithm" (SDA). Features that never cause the split of a set are redundant.
     This works but is unsatisfactory, because the order in which the features are used influences the form of the resulting tree, and ridiculous results can be obtained. The solution is to order beforehand the features in a Contrastive Hierarchy, and start with the most contrastive one. Given the hierarchy, only one tree results, and given the correct Contrastive Hierarchy, the correct tree results. [DG: The paper solves only half the problem, in that it is not explained how to obtain the correct Contrastive Hierarchy.]
     [DG: I have my doubts. I do not believe that people distinguish phonemes by minimal features: people need redundancy in their language, including in their phonemes. From the other papers on the subject it seems that the hierarchy is mostly determined so as to obtain the intuitively desired tree. The author's YouTube lecture of 28.10.2020, 19h00 did not improve matters; it does not address the question of how to obtain the elusive hierarchy, and his "explanation" of umlaut in German, with its "late phonetic rules outside phonology", and {round} as an "enhancement feature" sounds as a cop-out. I think the Contrastive Hierarchy is a straightjacket and CHT an exercise in minimality.]
     Illustrative examples of phoneme sets of various languages and the results obtained for them by the three algorithms presented here (Pairwise, SDA, Contrastive Hierarchical SDA) are provided.

* Ragagnin, E., Turco-Mongolic Relations: The Case of Particles, 2009, pp. 13.
In the Khövsgöl region in Mongolia several groups of people live in reasonable proximity of each other, some speaking a Turkish, others speaking a Mongolian language. The author reports on field work among the Dukhan (Turkic) and the Darkhat (Mongolic) speaking people. In this paper the particles iyǝk (from Dukhan) and asən (from Darkhat), both meaning `it really is (I've seen it myself)', are compared.

* Werner, H., Zur Etymologie des Namens Sibir `Sibirien', 2009, pp. 15.
Information-rich but answers-poor paper, describing, and sometimes supporting, many etymologies that have been suggested. One frequent suggestion derives the name from Sabir, a north-Caucasian tribe which moved to beyond the river Irtysh. Another splits the name as Sib-Ir and derives the -Ir from the name of that river; but correspondences of fewer than 3 phonemes are suspect.

* Bengtson, J.D., The Languages of Northern Eurasia: Inference to the Best Explanation, 2008, pp. 22.


* Bentley, J.R., The Search for the Language of Yamatai, 2008, pp. 43.
Using new information about the pronunciation of 3rd century Chinese, the author (re)examines place names from the 3rd century Wei Chronicles from Yamatai (Wa), and concludes that from the 31 place names, 17 suggest Japonic descent, 6 Ainu, and 8 miscellaneous/unknown.

* Dybo, A.V., Starostin, G.S., In Defense of the Comparative Method,or The End of the Vovin Controversy, 2008, pp. 140.
Rebuttal of Vovin "The end of the Altaic controversy" (2005).
The authors thank Vovin for the convenient list of all the criticisms to EDAL. They answer the criticisms in Vovin (2005) one by one in great detail, which accounts for the 140 pages. Some examples:
• The demolishing, in 24 pages, of 8 etymologies by Vovin (2005) is answered in 15 pages.
• To the criticism that EDAL is built on dictionaries and ignores the underlying research, the authors answer that introducing the underlying research would have made their dictionary into a "A Handbook on the History of the Altaic Debate".
• In answer to the criticism that EDAL sometimes ignores etymologies from the daughter languages and shows Proto-Altaic etymologies instead, the authors show that the in those cases the Proto-Altaic etymology is better in that it requires fewer arbitrary assumptions.
     The authors admit that their etymologies are not always regular and do not always explain the whole word, but point out that that is true for all traditional comparative linguistic work. The authors decline responsibility for the price of EDAL, and point out that the etymological dictionaries on which it is based as accessible free of cherge at The Tower of Babel Project (

* Ciancaglini, C.A., How To Prove Genetic Relationships Among Languages: The Cases of Japanese and Korean, 2008, pp. 36.
The comparative method applies only under very specific conditions; Japanese and Korean (and the other Altaic proto-languages) do not fulfil these conditions. Ergo, the genetic relationship between them, if it exists, cannot be determined by the comparative method.
     Some conditions are:
1. Because the method works on the sounds of words, the languages to be compared must have long histories of phonetic information. Reliable phonetic info for Korean and Japanese is only available for perhaps half a dozen centuries; for Proto-Indo-European the number is more like 3600 years.
2. The comparative method can go back in time for at most 3000 years. Starostin (and others) claim that the time depth of Altaic is about 7000 years, completely out of reach for the comparative method.
3. The comparative method applies to words only, not to structural types. The most striking similarities between Korean and Japanese are structural, but they are not amenable to the comparative method.
4. The comparative method is negatively sensitive to loans. Korean and Japanese have been in prolonged contact around the Bay of Bohay and in Korea, so extensive loans can be expected, polluting the comparative method, in spite of all efforts to remove the loans.
5. The comparative method has little defence against accidental matches, and they are quite common, f.e. Eng. `much' is not related to Sp. `mucho'.
6. The comparative method works only on languages with "internally articulated signs" and does not work on languages with "fixed signs"; "the older attested languages" have internally articulated signs, and Korean and Japanese have not.
[DG: I have no idea what this means. Latin, Greek, and Sanskrit words consist of stems + endings, and so do Korean and Japanese words (unlike Semitic words, and yet the comparative method works for these languages).]
     The above points apply also to morphemes. Proposed morphemes common to the Altaic languages, are unsystematic and far less convincing than those of PIE.
     The comparative method tends to yield information-rich sound laws, f.e. a law in which PIE k' corresponds to k, h, š, θ, and s in the various daughter languages. The proposed Altaic sound laws, however, are flat, information-poor, f.e. Proto-Altaic initial k corresponds to k in all daughter languages. This suggests that related pairs in Altaic were selected for their surface similarity rather than for their deep relationship, in other words, that they are of the type `much' ~ `mucho'.

* Robbeets, M., If Japanese Is Altaic,How Can It Be So Simple?, 2008, pp. 32.
Japanese differs typologically from the other four (Macro-)Altaic languages in that it only allows open syllables (where syllables ending in n count as open), and it is interesting to see how that has come to be. To solve this question, the author presents a study of consonant clusters in the Macro-Altaic languages, concentrating on medial clusters; initial and final clusters are rare in Micro-Altaic, although not in Middle Korean.
     It turns out that medial clusters in Micro-Altaic usually have the shape [nmlr][PTK] or more rarely [PTK][nmlr] [DG: as simple eye-balling will confirm]. The author then shows that Micro-Altaic medial clusters of the form n[PTK] turns into [bdg] in Japanese, thus simplifying many medial clusters and contributing to the number of open syllables. [DG: This is only a partial answer to the question in the title. It would be interesting to see what became of the other cluster patterns.]

* Kempf, B., Review of EDAL, 2008, pp. 6.
Critical review of the EDAL.
1. There are no good reliable etymological dictionaries of the component languages, so it is impossible to create a reliable etymological dictionary of a hypothetical parent.
2. Nursery words and onomatopoeia are used in etymologies.
3. Opinions of other scholars are brushed off.
4. Many sound changes in many etymologies do not conform to the proposed sound laws.
5. Etymologies are given for entities that were not present where proto-Altaic would be spoken (elephants, etc).
6. Great semantic leeway is used; e.g. `wilderness', `meadow', and `mountain' are considered equivalent for etymological purposes.
     For the future, the author suggests that from the present etymology set of Turkic and with improved etymology sets of Mongolian and Tungusic it should be possible to prove the existence of proto-Altaic, if it exists. And only if it exists, attempts should be made to add Korean and Japanese.

* Rybatzki, V., Middle Mongolic: When and Where was it Spoken?, 2008, pp. 7.
The paper consists of three loosely related and not well separated sections.
1. Traditionally "Written Mongolian" is distinguished from spoken Mongolian to the extent that different grammars are produced for both. Using a stanza from `A Treasury of Aphoristic Jewels' the author demonstrates that the difference is an artifact of the recording methods (hPags-pa, Uyghur runes, Sinitic characters, or Arabic).
2. The whereabouts of the speaker of Mongolian languages are followed from the 2nd century AZ, when the Mongolians lived in the Xiongnu empire in Eastern Mongolia, until the present. Much detailed information is given.
3. The history of the Mongol languages is divided in 6 eras, again from the 2nd century AZ till the present. For each era, the languages about which there is information are named.

* Parkvall, M., Which Parts of Language are the Most Stable?, 2008, pp. 17.
A language can obtain a feature in two ways: it can inherit it from the parent language or it can borrow it from a neighbouring language. [DG: It can also obtain it through internal development, but that is not considered here.] The paper tries to determine for a large number of features the ratio between these two influences.
     Formulas are designed for the correlation of a given feature F inside a known or purported language family, based on Greenberg's diversity index, and the same is done for areas. [DG: It is not explained how these areas are delineated.] The language family values are then averaged over all Languages in the WALS and the area values are averaged over all areas. The ratio of these two values is a measure for the stability of the feature in languages and its resistance against areal influence.
     The paper contains a list of the top ten WALS features with the highest family/area ratio; these are presumably the most genealogically stable features. High among the most stable features are: predicative adjectives behaving like verbs, and nominal plurals formed by suffixing. A similar list of the least stable features conain for example: having the same word for green and blue, and having demonstratives with distance contrast.
     The process was the repeated with the uncontroversial language families only. This results in detailed shifts in the list, which are all interpreted in the paper.
     There is also a table of the age of language families versus their internal diversity. Unsurprisingly the trend is: the older the more internally diverse. The last table shows the internal diversity of a number of disputed (macro-)families, where Amerind, Papuan, and Australian score about the same as random sets of languages, and Altaic (apparently macro-Altaic) and some fake language families (e.g. Penutian + Dravidian) do even worse.

* Nichols, J., Proving Dene-Yeniseian Genealogical Relatedness, (slides), 2008, pp. 36.
Considering how relatedness can be proved without doing reconstruction, the author claims that it can be done by showing that the number of "resemblant elements" significantly exceeds what would be expected by chance.
     The first 15 slides contain a crash course in statistics and Dolgopolsky mapping. This knowledge is then applied in a test for the relatedness of the East and West Caucasian languages, based on Nikolayev & Starostin's "North Caucasian Etymological Dictionary". The conclusion is that the number of "resemblant elements" is totally insuffucuent. Next it is applied to Ruhlen 1998, "The origin of the Na-Dené", comparing Yeniseian and Na-Dené, and again the verdict is "insufficient evidence".
     The author now turns to morphological paradigms, starting with the pronominal affixes of Wiyot and Yurok (related, p < 10^−7). Next the 3 TAM prefixes of Yeniseian and Na-Dené from Vajda's 2008 "A Siberian Link with Na-Dené Languages" were compared, first each separately, which yielded insufficient evidence, and then adding the information that the 3 TAM morphemes occur in the same order in both families, which reduces the probably of chance similarity to < 10^−5, which is near-sufficient. The probability that 2 shape prefixes are equal and one is almost equal by accident is also < 10^−5. The probability that these two unlikely phenomena occur in the same two families is so low that the author concludes that "Yeniseian and Na-Dené are related". This conclusion is then further supported somewhat by an analysis of the lexical comparanda from Vajda (2008).
     [DG: Not all slides were understandable for lack of the accompanying lecture.]

* Vajda, E., A Siberian Link with Na-Dene Languages, 2008, pp. 60.
The author argues that when two language families are found to be related, our knowledge of both families should benefit: problems in one could be cleared up by data from the other and vice versa. Such interaction boosts the confidence in the relationship. The author sets out to show this effect in the proposed relationship between the Yeniseian languages of Siberia and the Na-Dené (Athabaskan-Eyak-Tlingit) languages of America.
     An obvious similarity between the Yeniseian and the Na-Dené languages is that both are heavily prefixing on the verb. However, so are Sumerian, Burushaski, Abkhaz, and Ha (Bantu). But the author shows that the order and the details of the prefix morphemes are very different for all these languages. So the fact that these languages are heavily prefixing is not significant; the order of the prefixes, however, is.
     The author shows that the prefix orders of Athabaskan, Eyak, and Tlingit differ only minimally and claims, without showing, that the prefix orders of other New World languages is completely different. (The author objects repeatedly to the idea that Haida would be in some way related to Na-Dené.)
     Putting the pre-proto-Yeniseian and the Na-Dené verbal templates side by side, the following correspondences in the slots can be observed (from left to right):
1. spatial morphemes, sometimes derived from nouns for body parts;
2. tense/aspect/mood (TAM) morphemes (in Yeniseian);
3. pronominal elements;
4. perfective/stative/resultative morphemes;
5. one or more pre-verb morphemes y, a velar (l, χ, q), and/or d;
6. the verb base;
7. tense/ascpet/mood (TAM) morphemes (in Na-Dené).
The TAM morphemes (l, n in Yeniseian; ɬ, ñ in Na-Dené) differ in position; the explanation is that in Yeniseian they pertain to (the remnants of) an auxiliary verb, and in Na-Dené they pertain to the main verb. "These similarities go far beyond typological coincidence." In the next 19 pages the Yeniseian and Na-Dené contents of these morpheme slots are compared in detail. Especially the detailed similarity of the order of the pre-verb morphemes is striking.
     Presently both the Yeniseian (i.e. Ket) and the Na-Dené languages have phonemic tones, but Yeniseian started out without tones because its present tones can all be associated with non-tonal features in Na-Dené. For example, Ket high tone is associated with a final glottal stop on the corresponding Na-Dené word. Many similar correlations are described, exemplifying mutual benefit.
     Lexical comparison between Yeniseian and Na-Dené is non-trivial, for the following reasons:
1. word ends were reduced in proto-Yeniseian, resulting in high or abrupt (glottalized) tones;
2. disyllable words collapsed, yielding falling and rising-falling tones;
3. Ket and its sister language Yugh do not always agree on tone, which gives additional information.
Only items of core vocabulary, notably body parts, natural phenomena, and basic actions appear to be cognate.
     The last 30 pages detail the sound rules; they specify for each Na-Dené phoneme + conditions the resulting Ket (and Yugh) phoneme(s), followed by examples. These rules are often ambiguous, i.e. one Na-Dené phoneme may correspond to more than one Ket phoneme.
     Finally the author poignantly comments how only a few decades ago both Native North American and Native North Siberian boarding school children were punished for speaking ancient words...

* Yeon Jaehoon, Is There Ergativity in Korean? The Definition of Ergativity and Other Uses of the Term 'ergative', 2007, pp. 16.
Mainly a lament about the misuse of the term `ergative' for all kinds of non-mainstream verbal constructions. After defining `ergativity' as "nominative markings for nouns used with some, mainly intransitive, verbs and ergative + accusative markings for nouns used with transitive verbs", and discussing some variants, the author shows that typical Korean constructions like `is visible [to me]' for `I see' and sentences with multiple subjects are ergative only under weird definitions of ergativity.

* Robbeets, M., Koguryo as a Missing Link, 2007, pp. 20.
Almost all Koguryo material comes from place names in Korea, but the problem is that we cannot be certain that they represent the language of the Koguryo kingdom [DG: if there was such a language at all; see Janhunen, (2005)].
     The cognate etymologies of Beckwith (2004), suggesting that the Koguryo language was Para-Japonic, are vetted by the same criteria as used in Robbeets' thesis (2005), where they a labeled Type I (inconsistent matching); Type U (universal word); Type B (borrowing from a third language); and Type S (semantically loose).
     Of the 80 cognate etymologies between A = Old Koguryo and B = Old Japanese, 15 were found to be of Type I, 3 U, 19 B, and 7 S, leaving 36 probable Japanese-Koguryo cognates. For 18 of these Korean etymologies too could be found.
     To confirm a genetic relationship between Koguryo and (Para-)Japonic the author tries to find regular sound correspondences, but is severely hindered by the dearth of material: even for the most obvious correspondence k ~ k, only 5 cognates can be found (2 of which also have Korean explanations).
     The 5 morphemes that Beckwith identified are all rejected: 3 I, 1 U, 0 B, 1 S, but 4 of them can be related to Korean.
     The conclusion is that the material suggests an "affinity" between Old Koguryo and Old Japanese, but the evidence is weaker than that between Old Korean and Old Japanese, as presented by the author (2005).
     [DG: I find the numerals (3, 5, 7, and 10) quite convincing. They seem to indicate that a people, speaking a (Para-)Japonic language, lived in (mainly west and south west) Korea long enough to insert their numerals into place names. This has consequences for the opportunities for language exchange between Korean and Japanese. Whether that (Para-)Japonic language was Koguryo-go is immaterial (it probably wasn't; it's common people who give names to places, not officials).]
     [DG: How could the Koguryo dynasty govern such a huge realm for more than seven centuries without leaving behind documents or even graffiti??]

* Robbeets, M., How the Actional Suffix Chain Connects Japanese to Altaic, 2007, pp. 60.
Points out that the Altaic languages are rich in derivational morphemes (suffixes) and that similarities between such suffixes in the 5 Altaic languages should strengthen the evidence for the inclusion of Japanese in Altaic. The interest is especially in the `actionality' suffixes; they modify the word to express such notions as diminutive, inchoative, resultative, iterative, momentaneous, and durative. Suffixes of this type occupy places very close to the root of the word and are therefore more resistant to being borrowed than more final suffixes with more incidental semantics. Six actionality suffixes are identified in Old Japanese, and these suffixes are then followed through the other Altaic languages. Of these 6 suffixes, 3 were found in all 5 branches of Altaic, and 3 were found in 4 branches.

* Robbeets, M., The Causative-passive in the Trans-Eurasian Languages, 2007, pp. 44.
Morphemes close to the stem are interesting in that they are more resistant to borrowing than other features. In the Altaic languages that position is taken by a causative-passive morpheme. Causativity and passivity are not mirror images: passivity does not change the situation, just the emphasis, but causativity changes the situation: it adds one participant. Causativity semantics often occurs together with permissivity semantics, but in many languages the subject of the original sentence appears as direct object when causativity is involved and as indirect object with permissivity.
     The stability of a morpheme is determined by three properties: how difficult it is to perceive it as a separate unit; how difficult it is to pronounce it in isolation; and how resistant it is to phonetic erosion.
     A strong case is made for the existence of a proto-Trans-Eurasian causative-passive morpheme *-ti- on the basis of comparative evidence:
the proto-Japanese causative-passive morpheme -*ta-;
the proto-Korean causative-passive morpheme -*ti-, where Middle Korean -chi- is derived from -ti-ki- `double causative' and -chwu- from -ti-kwu- `id.';
the proto-Tungusic causative-passive morpheme -*ti-, with Manchu *-tA-, Udehe -si-, etc.;
the proto-Mongolian causative morpheme -*ti-. which interacts with the medial morpheme -ra-; and
the proto-Turkic causative-passive morpheme -*ti-.
Each of these is amply supported with examples.
     The a in the proto-Japanese -*ta- is probably due to analogy with other proto-Japanese morphemes and particles. The original meaning of the morpheme was causative, but the passive meaning was probably already present in the proto-language. The passive meaning may have been pushed out in Mongolian by the medial morpheme. This may point to a development starting from the center.
     On page 12 the author introduces the following interesting and ingenious piece of reasoning. In Middle Korean monosyllabic verb roots with high tone usually lose their tone when an ending follows, but keep it in the extended stem/infinitive, formed with the "ending" -ㅏ/-ㅓ (-a/-e) (from an older -ㅏ (a) through vowel harmony), which suggests that the -a/-e is not an ending but rather an auxiliary verb stem. Now the monosyllabic verb stem et- `to obtain' has rising tone in Middle Korean, which suggests that it originally had two syllables. The author reconstructs it as *a.ti-, which with -ti- as causative particle, would give the stem a- the meaning `to have'. This is then compared to the proto-Japanese stem a- `to exist'.
[DG: Indeed the difference between pat- `to receive' and et- `to obtain' is that with pat- somebody else gives it, and with et- the speaker him/herself causes it to be acquired. So et- has a much higher degree of causativity. If a- indeed means `to exist, to be there', it makes sense that, when asked "What are you doing?", the addressee answers meg-a `eating is there' = 'am eating'; this explains its use as a sentence ender. Explaining its use as a converb is somewhat less obvious but not hard: meg-e daesseyo `eating being there, he applied' = `he stuffed himself with food'.]
[>>>>DG: Unfortunately this does not work: et- cannot come from a.ti-, which would yield aet-. The Yale `e' is not an /e/. Robbeets answered (2022-04-07): "Yale `e' was /e/ in the 15-th century (Korean vowel shift)". Still /a/ would umlaut to /ae/ rather than /e/ (unless no /ae/ was available in the language).]
     The second part of the paper is concerned with the "ancillar auxiliary proto-Trans-Eurasian *ki- ‘do’". [DG: There is no word `ancillar'; on the other hand, words mean what you pay them to mean.] Early in the history of the Altaic languages this auxiliary verb has been grammaticalized into a morpheme *-ki-.
In Japanese it is the source of the vowel which, for example, turns kura- ‘(be) dark’ into kure- ‘get dark’. This requires the k to be dropped, which is supported by the relation takai ~ takaku.
In Korean it appears as -ki-, -hi-, or -i-.
Its original meaning is still apparent in Evenki and Udehe: -ŋ- ~ -ŋi:- ‘make, build, create', for example xokto ‘road’ => xoktoŋi- ‘make a road’. But it also exists as a causative morpheme in Tungusic.
Proto-Mongolian and all Mongolian languages have a verb ki- `to do, make'. Remarkably the author does not speculate whether this verb is the origin of the morpheme.
Proto-Turkic had a verb *kïl- ‘do, make’, but its exact relation to the causative morpheme *-ki- is unclear. Again all this is amply supported by evidence.
     On the basis of detailed comparison between the supporting examples, the author rules out coincidence and universal linguistic principles, and based on eleven considerations rules the similarities to be of genetic origin. Some of the considerations are:
• The morphemes discussed here immediately follow the stem, in a position that is very resistant to borrowing.
• When morphemes are borrowed, they are sometimes/often borrowed in adjacent pairs; there is no sign of that here.
• The morphemes occur in all 5 branches; borrowing by 4 language families seems highly unlikely.

* Vovin, A., Korean Loanwords in Jurchen and Manchu, 2007, pp. 12.
Three words from Vovin `Why Manchu and Jurchen Look So Non-Tungusic?' (2006) plus four new words are analyzed in depth. An example is Manchu misu-n `miso'. Tradition has it that this word was borrowed from Japanese, which had borrowed it from Korean Early Middle Korean mico, on the grounds that if it was borrowed directly from Korean the c would have been preserved. But the author presents several words in which c from other sources is represented by s in M/J.
     The author draws two conclusions:
There is early influence of a Koreanic language on M/J, which can probably be attributed to the elite language of Koguryo or Paekche.
There is no need to assume occasional Japanese mediation in this influence.

* Vovin, A., Once Again on the Etymology of the Title `qaγan', 2007, pp. 11.
After much deliberation the author analyzes the word qaγan (previously attributed to the Ruan-ruan language (Vovin 2004)) as qa-γa-n, with qa from proto-Yeniseian *qεɂ `big', γa from proto-Yeniseian *qʌ:j or *χʌ:j `ruler', and -n from Mongolian for singular. This is confirmed by dar-Ga: `ruler of the North', where dar means `North', and by the plural qaγad in Mongolian.
     [DG: I am confused. With the reversal of meaning between qa and γa, qa-n is no longer explained (`big-noun'??), and neither is qa-tu-n (`big-female-noun??). The author does not address this.]

* Vovin, A., Once Again on the Tabgač Language, 2007, pp. 16.
It is possible that the name `Tabγač' is the same as `Chuvash', but that does not tell us anything about the language of the Toba-Wei (5th century AZ in Northern China). Eleven words were known, written in Chinese characters. Reading these characters according to Pulleyblank (1991), and taking into account which syllables could not be expressed in them, the author reconstructs 9 para-Mongolic words (e.g. čino `wolf' ~ Modern Mongolian чоно), and 2 Turkish loans into Tabγač; some of them have para-Mongolian suffixes.
     Two more words have come to light later. One is immediately clear: Tabγač nyak kan 'dog' ~ Khitan *ńaxa ~ Middle Mongolian noqai ~ Modern Mongolian нохой. The second is more problematic: moqolo `bald', because there is no known Mongolian equivalent. But there is Middle Mongolian muqular `hornless' and through a complicated reasoning involving Tungusic loans the author connects this to Tabγač moqolo `bald'.
     The conclusion is that there are strong indications that Tabγač is a para-Mongolic language (like Khitan).

* Saitô Yoshio, The Vowel System of West Middle Mongol, Mongolian Studies, vol. 29, 2007, pp. 141-145.
As represented in the `Muqaddimat al-Adab', in Arabic transcription. Although the Arabic transcription of Middle Mongolian did not distinguish between o, ö, u, and ü, and represented all four of them by a waw, it is sometimes possible to discern their presence by the choice of the surrounding consonants (s/s-dot, k/q). Using this technique the author finds eight vowels: high: i, ü, ï, u, middle: e, ö, o, low: a, with length distinguished. Vowel reduction also occurred: the word čagan `white' was recorded as jagan, jygan, and jgan.

* Vovin, A., Why Manchu and Jurchen Look So Non-Tungusic?, 2006, pp. 12.
Although genetic material shows unequivocally that Manchu and Jurchen belong to the Southern branch of the Tungusic languages and do not form a separate branch by themselves; peculiarities in their vocabulary, morphology and syntax make them stand out as quite different. It is likely that these differences are the result of intensive contact with other, non-Tungusic, language(s).
     The obvious choice for an influencing language is Mongolian, and indeed a number of (para-)Mongolic loans in Manchu/Jurchen have been identified (Janhunen). However, these do not cover all words in M/J for which there is no Tungusic etymology, and the author argues that another influencing language was a para-Korean language. This is based on 13 such words that match almost but not exactly Middle Korean words. An example is Manchu fulehe `root', compared to MK pulhuy' `id.'. Also, Manchu/Jurchen has a genitive case, unlike the other Tungusic languages. It uses the ending -i, which cannot have been borrowed from Mongolian, which has -n, but is a match for MK -uy.
     Even borrowing from para-Korean does not explain all non-Tungusic elements in Manchu/Jurchen, and 14 words are presented that must have been borrowed from an unknown language. Remarkably these are all basic words: `warm', `cold', `to see', `woman', etc.
     This proves that 1. a para-Korean language was present in Koguryo; and 2. at least one unknown language was in the region in a position to have close contact with Manchu/Jurchen.

* Martin, S.E., A Reference Grammar of Korean, Tuttle Publishing, 2006, pp. 1056.
This book provides tons of information not available elsewhere in English; one may quibble about the presentation, but that fact stands.
     Although it is a very thick book with small margins it actually consists of four parts:
1. a grammar of 347 pages, detailing all the ins and outs of the structure of Korean (and there's a lot of it);
2. an appendix of 67 pages, listing, for example, all verb stems ending in -ㄹ, and many other word classes;
3. a grammatical lexicon of 542 pages, listing dictionary-wise all forms and endings, with translation, examples, and historical background; and
4. an index of 76 pages.
As I said, encyclopedic.
     Its use of (Yale) romanization instead of Hangeul has been reviled by many reviewers and it is indeed an eye sore. But the present-day Hangeul symbol set was designed for modern Korean, and cannot represent many Old and Middle Korean syllables. Scribes of old were not restricted to a fixed set of consonant combinations, and could write on their paper any combination of double or even triple patchim if they needed them. (Recently a Unicode set for Old and Middle Korean has been established, the Jamo set.) Using the romanization, however ugly, allowed the author to freely write down any historical, hypothetical, or partly incorrect form he needed for his explanations. For example, the stem of 돕다 – “to help” is “tow” but that cannot be written in Hangeul. So first I was annoyed by the ugly romanization (the use of “e” for a sound that is more like the “o” in “hot” still irks me no end) but I'm beginning to see the usefulness of the Latin alphabet over the Hangeul one for scientific linguistic purposes.
     Another thing that kind of annoyed me is that the author describes the “speech of the educated Seoulite” rather than “Standard South Korean”. They differ in subtle points: the 여 is mostly pronounced as 에, producing f.e. 멫 for 몇; the 예 is pronounced 이, which makes 예쁘다 into 이쁘다; and the infinitive has -아 only after 오, so we find 받어요 for 받아요. Oh well, anybody who can read this book will get over that.
     I liked very much that the author has consistently indicated which syllables and endings cause tensing of the following consonant when applicable. Again Hangeul does not shows this, but you still have to know it when you want to learn the language properly.
     Definitely worth every Eurocent.

* Finley, S., Vowel Harmony in Korean and Morpheme Correspondence, 2006, pp. 15.
In spite of the title, an in-depth analysis and explanation of the vowel harmony in Korean sound-symbolic words.
     Author's conclusion: "The use of faithfulness-based correspondence constraints and multiple features allows for a complicated and seemingly idiosyncratic harmony system to be explained in a relatively straightforward manner. The violation of the Harmony/Inventory Theorem caused by the unusual behavior of high vowels makes sense in the context of the morphological and multi-featural nature of harmony in Korean ideophones."

* Janhunen, J., The Lost Languages of Koguryŏ, 2005, pp. 20.
Koguryo, covering central Korea, Manchuria and the Liaodong peninsula, was made up of several ethnic groups speaking several different languages; there was no "Koguryo language". Chinese was the prestige language and the language of administration.
     Pre-Old-Korean was the language of Silla. It was initially spoken by relatively few people and thus quite uniform, and its expansion resulted in a uniform almost dialect-free Korean. Given the multi-lingual nature of Koguryo and Paekche, Shilla'sexpansion must have extinguished several local languages.
     The language of Paekche was Para-Japonic, as is suggested by the Old-Koguryo place names. Para-Japonic is the only language next to Old Korean and Chinese that has been documented in Korea. In the 4th century BZ its speakers started to move to the Japanese islands, taking their language and the Yayoi culture with them.
     Although there was no nation-wide Koguryo language, there was a Koguryo dynastic language. The territory of Koguryo was bordered by speakers of Mongolic (west of the Liao river), Tungusic (east of it) and Amuric (in central or southern Manchuria). At present Amuric Gilyak is the least prominent language in the region, but the author points out that Gilyak has its own words for several metals, among which `gold' and `silver', which shows that it has seen better times: some of the purported loans from Manchu to Gilyak, for example higher numerals, may actually be loans in the reverse direction. The most probable hypothesis for the dynastic language of Koguryo would be Tungusic, because the other candidates were already state languages of other states (Amuric with Puyo).
     Korean has a Para-Japonic substratum, and Japonic has a Korean adstratum from its coexistence on the Korean peninsula. Traces of Korean's coexistence with Manchu are less specific.
     Unlike the other regional languages, Japanese was not originally from the region but stems from the Yangtse Basin (Janhunen, 1999). The author presents four scenarios for the trek from the Yangtse basin to Japan, and finds the one from Shandong to Liaodong to Korea the most probable, but does not exclude the others.

* Unger, J.M., When was Korean First Spoken in Southeastern Korea?, 2005, pp. 18.
After considering several published theories, the lack of linguistic diversity both in Korea and Japan, and the archeological evidence, the author concludes that the language of the Yayoi was either a creolized form of a language long ago separated from Korean [DG: creolized with what other language and under what pressure?], or an unrelated language originating from the Jiangnan region. This language was spoken on the Korean peninsula until the 4th century AZ, when the Yayoi were pushed overseas by the advancing Koreans, who then founded the Three Kingdoms. A corollary is that the Three Kingdoms all spoke forms of pre-Old-Korean.

* Vovin, A., Koguryo and Paekche: Different Languages or Dialects of Old Korean?, 2005, pp. 33.
Some place names on the territory of the ancient Koguryo and Paekche realms look like Old Japanese (e.g. Beckwith, 2004), but that only shows that Old Japanese was once spoken there, but not that Old Japanese was the language of Koguryo or Paekche.
     In the first seven centuries of the first millennium the kingdom of Koguryo was the most powerful entity in the region and it must have exerted large influence on its western and northern neighbours the Jurchen/Manchu, leading to the introduction of Koguryo loan words. The author cites 13 Jurchen/Manchu words that have no etymology inside Jurchen/Manchu, but can easily be explained as loans from a Korean language. Some Korean-like influence in the morphology is also detected.
     For Paekche the author turns to words in Old Japanese texts that are identified in the text as being from Paekche. For 18 of these words a relation to a Korean language is very likely, for example Paekche sitoro (シトロ) `belt' → Middle Korean stuy → Kor. ddae 때 `belt'.
     The conclusion is that the languages of Koguryo and Paekche were Korean dialects.

* Toh SooHee, About Early Paekche Language Mistaken as Being Koguryo language, 2005, pp. 22.
The Samguk-Sagi was written in 1145 by Kim Pushik 金富軾 (1075-1151), but the renaming lists, ordered by King Kyŏngdŏk of Silla (景德王; ruled 742–765) are from 757. Examining the campaigns and conquests of king Chinhŭng (mid 6-th century) the author concludes that the middle part of the Korean peninsula, which is where the majority these place names hail from, did not belong to Koguryo, but to Paekche, and that the language extracted from these names was Early Paekche.
     Place names in the real (non-Paekche) Koguryo area are also discussed.

* Robbeets, M., Is Japanese related to Korean,Tungusic,Mongolic and Turkic?, (thesis), 2005, pp. 445+523.
The idea is to collect all etymologies of Japanese words that have been proposed in the literature and that relate them to Altaic (Korean, Tungusic, Mongolic and Turkic); weed out unreliable ones; and draw conclusions from the remainder.
     There are 4 reasons to reject an etymology when used for the purpose of arguing for genetic relationship between two languages, A and B:
Type I: a segment of the word in A is matched with the wrong segment of the word in B;
Type U: the word is universal;
Type B: it is more likely that the common word was borrowed by both A and B from a third language;
Type S: the semantic correspondence is too loose.
(The labelling of the criteria with Type I, etc., is from Robbeets, `Koguryo as a Missing Link' (2007) and does not yet appear in the thesis.)
     The hypothesis is tested using linguistics, archeology, anthropology, molecular biology. After three introductory chapters, chapter 4 gives a meticulous, argued description of the method(s) used to answer the question. Chapters 5 to 9 do the weeding. Of the 2055 Japanese-Altaic etymologies, 635 survive, and in chapter 10 attempts are made to find sound correspondences. The sound correspondences fully explain 359 etymologies. There are also 14 morpheme etymologies. This allows the question in the title to be answered with Yes.

* Vovin, A., The End of the Altaic Controversy, 2005, pp. 62.
Personal account why the author no longer supports the Altaic hypothesis. The basic reason is that the closer one looks at the evidence the more it disappears.
     The EDAL ("Etymological Dictionary of the Altaic Languages", Starostin et al., 2003) should be a monument of proof for the Altaic hypothesis, but it is found sorely lacking: it is full of inconsistent or unique sound correspondences; it is based almost exclusively on dictionaries and ignores the underlying research; words are segmented into morphemes in Proto-Altaic in ways that are inconsistentd with known etymologies in the daughter languages; etc. The author selects 8 etymologies that seem to have great proving power because they include words from all five branches and use non-trivial sound correspondences. Next each of these etyma is demolished because of the above deficiencies and various other grounds, e.g semantics that did not exist at the time of Proto-Altaic (e.g. "Buddhist temple"), incorrect etymology in a daughter language, non-existing word, etc. Next the authors shows that no regular sound laws can be constructed for the Proto-Altaic reconstructed diphthong yo in first syllable, even if the context is taken into account. The conclusion is that "EDAL's Altaic is a construct of the human mind", and as such ends the Altaic controversy. See, however, Dybo & G. Starostin(2008) for a rebuttal.
     As an aside, the author implies that similar criticism to a lesser extent also applies to "A North Caucasian Etymological Dictionary" by S. Starostin et al.

* Robbeets, M., Does Doerfer's Zufall Mean `cognate'? The Case of the Initial Velar Correspondence in Altaic, 2005, pp. 34.
We are deep in the fray of the Altaic debate here: Stefan Georg calls the question "a breathtakingly and arrogantly slanderous remark on Doerfer" in his review of Robbeets' thesis. Both phrases seem a bit over the top.
     When Doerfer in 1963 examined the 64 etymologies demonstrating Altaic *g- ~ Mon. g- ~ Tk. q-, which Ramstedt has has found in 1949, he rejected 63 of them as unreliable, leaving only one acceptable etymology: Mon. γar `hand, arm' ~ Tk. qar `arm'. And since one etymology proves nothing, Doerfer declared this etymology "Zufall" `coincidence'.
[DG: The `die angebilde Gleichung' in the German quote from Doerfer's book should be `die angebliche Gleichung' `the alleged equation'. Bad OCR?]
     Today we have the EDAL with around 1500 Macro-Altaic etymologies and Robbeets' thesis with around 350 vetted etymologies that include Japanese, and it is interesting to see if now more than one etymology for Altaic *g- ~ Mon. g- ~ Tk. q- can be found.
     The author finds 69 good etymologies with an initial velar, for which the CVC-s correspond. Among them is Doerfer's etymology: pAlt kata ~ pJ *kata ‘shoulder’ ~ pTg *gāla ‘hand’ ~ pMo *gar ‘hand’ ~ pTk *karɨ ‘arm’. This means that for Doerfer's etymology and the other 68 "Zufall" can be ruled out "with a considerable degree of probability".
     Out of the 69 etymologies 31 point at a pAlt *k, 13 point at pAlt *g, and 25 are undecided. In the 13 etymologies with pAlt *g the second consonant in the CVC triple is always voiced, which suggests the possibility that pAlt *g is the result of voicing of pAlt *k in certain environments.

* Miller, R.A., Turkic böz `fabric' in Korea and Japan, 2005, pp. 9.
Inaccessible; relevant points reconstructed from fragments in Vovin (2018, `Fabrication ...').
     The author reacts to Georg (2003) `Japanese ...', in which Georg adduced the Chuvash word pir ~ Common Turkic böz as proof of the existence of zr (rhotacism) in Chuvash, because the word was originally a loan from Arabic (bazz) and had a z [DG: and if there was rhotacism in Chuvash the words had z in proto-Turkic and cannot support the Altaic hypothesis]. The author points out that there is another word in Chuvash püs `calico', very likely borrowed from the same source, which escaped rhotacism. This casts serious doubts on the existence of rhotacism in Chuvash [DG: which would save the words as support for the Altaic hypothesis].
     The word is a Wanderwort, also occurring in Middle Mongolian as böz `cotton textile', and in Manchu as boso `fabric'. It is even recognizable in Japanese pusuma `bedding, coverlet', in which -ma is a suffix meaning `made of...'.

* de la Vaissière, É., Huns et Xiongnu, (in French: Huns and Hiung-Nu), 2005, pp. 24.
All Western sources that encountered the Huns called them "Huns": their name is never in doubt. In Sogdian they were called "Hwn". They were assorted tribes from the foothills of the Altai mountains, unified under Hiung-Na culture. The paper does nor address language issues.

* Robbeets, M., The Classification of the Japanese Language: Belief or Argument?, 2004, pp. 8.
The author points out that the comparative method can help decide whether two languages are genetically related, but that is is no help in selecting the two languages; for that, external information is required.
     Referring to Robbeets (2003, thesis), the author sets out to subject the cognate etymologies presented there to stricter criteria. More in particular, cognate etymologies are rejected when they fall in the categories that are described more clearly in Robbeets `Koguryo as a Missing Link' (2007): Type I: internal conflict; Type U: universal word; Type B: more likely borrowed; Type S: semantically loose.
     The 1804 cognate etymologies from Robbeets (2003) are selected that involve Japanese. Of these, 1229 are rejected: Type I: 474; U: 89; B: 181; S: 485. This leaves 577 etymologies. To prove genetic relationships one needs sound correspondences, and to obtain these a bottom-up algorithm is applied to the 577 etymologies. This resulted in a number of sound laws (not included in the paper). Not all etymologies were fully compatible with the sound laws but 349 core etymologies were; 42 of these concern basic words on the Swadesh 100 list. Seven of them are demonstrated in detail; 3 involve Korean (one of them through metathesis); 4 involve Tungusic; 6 involve Mongolic; and 6 involve Turkic.
     Morphology show a similar but weaker image.
     The author answers the question "Is Japanese Altaic?" with "yes" because
1. the existence of triple-phoneme correspondences rules out chance similarity;
2. the sheer number of cognate etymologies (349);
3. the number of branches involved (5);
4. the overall similarity in the phonetic systems;
5. the fact that the core etymologies cover almost half of the Swadesh 100 list.

* Ramsey, S.R., Accent, Liquids, and the Search for a Common Origin for Korean and Japanese, 2004, pp. 12.
The paper assesses the progress in research on the relationship between Korean and Japanese since the work of Martin's `Lexical Evidence Relating ...' (1966).
     Korean and Japanese have been compared since around 1900 for two reasons: 1. the languages are structurally uncannily similar; 2. the countries are close together. But are these reasons sufficient? It is worrying that as we go back in time, Korean and Japanese seem neither to diverge nor to converge.
     Accent pattern in some parts of Korea are very similar or equal to those in some parts of Japan: e.g. the accent system of Hamgyeong Korean is the same as that of Tokyo Japanese: the accented syllable is followed by a sharp drop in tone. But then, Thai and Chinese have virtually identical tone systems (which still misled linguists up to the middle of the 20th century), any yet they are not related. Also, the accent patterns in Japanese are ancient, at least dating from the Nara period (mid 8th century AZ), whereas the Korean accents appear to have arisen from the segmental structure of the language around the 15th century, as shown in Ramsey `Proto-Korean and the Origin ...' (1991). This means that the accent patterns cannot be used in support of a genetic relationship of Korean and Japanese.
     The fact that neither Korean nor Japanese distinguishes l from r is often used to support the claim that Korean and Japanese are related. There are no signs that Japanese ever had more than one lateral, but Korean probably had. In Middle Korean the tones of monosyllabic verb stems were usually predictable from the final consonant (Ramsey `Some preliminaries to ...' (1996)), but there is a minimal pair which differs in tone only: kol- L `to exchange' and kol- H `to grind'. This suggests that the ls were different: kol¹- L and kol²- H. Another glimpse of two different ls can be gathered from the use of the causative particle -hi-. This particle takes different shapes determined by the preceding consonant, but with verb stems that end in -l it takes the forms -i- or -Gi-, without evident rule. Again the verbs may end in different ls, with causing the -hi- to change to -Gi-, and causing the h to be lost. Finally there is the fact that the l in the noun object marker -ul and the l in the verbal future marker -ul are consistently spelled with different Chinese characters, which shows that they were different. Because the future marker causes tensing of the following consonant (if applicable) and because it is sometimes written as ㄹㆆ (rieul + glottal stop), the author suggests it was a lateral obstruent, and identifies it with l¹; l² is then a lateral sonorant, most likely just l. This means that the fact that both modern Korean not modern Japanese have only one lateral cannot be used in support of the genetic relationship between those languages.

* Vovin, A., Some Thoughts on the Origins of the Old Turkic 12-Year Animal Cycle, 2004, pp. 8.
Many Asian cultures divide the years according to the 12-Year Animal Cycle (rat, ox, ..., dog, pig). In Chinese and Southeast Asian languages the animals in the Cycle have traditional names, which are different from the common words for the animals (the traditional names may stem from a Katuic (Viet-muong) language). The Central, East, and Northeast Asian languages, however, use the common words for the names of the years in the 12-year cycle, except that in some languages one or two names are borrowed from a culturally more prestigious language.
     This effect is the strongest in Old Turkish, where 5 out of the 12 names are non-Turkish: pig, dragon, rat, ox, and horse. The name for the dragon, luu, can be identified as borrowed from Chinese, but the donor language [DG: or donor languages?] of the other four cannot readily be identified. Excluding all other possibilities the author concludes that it is highly likely that they are from the Ruan-ruan language.
     The Ruan-ruan (Ruan-Yuan, Roran, etc.) language was the language of the Ruan-ruan Khaganate (330-555 AZ) in Central Asia. Only 4 words from this language are known, and the form a nice set: qa-n `khan', qa-tu-n `khan's wife', qa-γa-n `emperor', and qa-γa-tu-n `empress', which allows us to identify three morphemes: -γa- `great', -tu- `female', and -n `noun'. These are not indicative of any Central Asian language.
     The 12-year cycle adds 4 possible words to this list. A ninth word is identified from another loan in Old Turkish. On the basis of this evidence the Ruan-ruan language seems unrelated to any known language in the region. (See, however, Vovin, `Interpretation of the Hüis Tolgoi Inscription', 2007).

* Vaux, B., [atr] and [back] harmony in the Altaic languages, 2004, pp. 17.
Simplifying the argument we can say that 1. vowel harmony in Mongolian and Tungusic is traditionally described as [front]-[back] harmony; 2. the i has a palatalizing effect on the preceding consonant; 3. this indicates that i is a [front] vowel; 4. the i is transparent to the vowel harmony process; 5. therefore the i must operate along a different axis than the vowel harmony process; 6. the i is defintely a [front] vowel, so the vowel harmony cannot be [front]-[back]; 7. [ATR] is a good alternative.
     Taking examples from many Mongolian and Tungusic languages and showing similar effects in a Scottish dialect (Buchan) and in Armenian, the author argues that proto-Altaic already had [ATR] harmony. It is then shown how the present-day Turkish [front]-[back] harmony system could easily have developed from it.
     The paper starts with computations on the F1 and F2 formants of Khalkha vowels, showing that F2-F1 as an indication of vowel backness gives contradictory results, as F2-F1(ü)=380 < F2-F1(u)=420, whereas F2-F1(ö)=530 > F2-F1(o)=440. These computations are shaky, however, because if we use F2 as an indication of vowel backness the contradiction disappears: F2(ü)=650 < F2(u)=850, and F2(ö)=930 < F2(o)=1060. These results can better be considered as part of the F2 versus F2-F1 debate. (Values estimated from Figure 4 in the paper.)

* Vajda, E.J., Ket, Lincom Europa, Languages of the World 204, Munich, 2004, pp. 99.
To do.

* Divitçioǧlu, S., Beyond the Hunnic Couplet, Central Asiatic Journal, vol. 48, #2, 2004, pp. 185-191.
Completely hypothetical alternative interpretation of the first line of the Hiung-Nu couplet. The author reads the line as "süke telegen", and translates the parts as `army', -ke `to', tele `a name meaning "cart"', and -gen `a honorific suffix'. The name `Tele' is illustrated by e.g. Pang Tele, an Uyghur chief (early 1200s). No over-all translation is given. [DG: This seems rather ad-hoc and does not fit the Old Chinese translation.]

* Beckwith, C.I., Koguryo: The Language of Japan's Continental Relatives, 2004, pp. 297.
The book argues that the language of Koguryŏ was a (para-)Japonic language. Much evidence is presented and many alternative hypotheses are shown to be untenable.
     Chapters 1 and 2 are introductory. Archeology shows us that the Yayoi first entered North Kyushu from south Korea in the 4th century BZ; there was no significant non-Yayoi immigration afterwards and further development was internal.
Chapter 3 analyses the available toponyms from the `Samguk Sagi'.
Chapter 4 concerns the Chinese phonetic values of the characters at the time they were used for the transcription of Koguryo forms. [DG: I don't understand why this is treated after Chapter 3, where it is used.]
Chapter 5 analyses the phonology Koguryo and its relation to Old Japanese, on the basis of the Koguryo/Japanese common pairs collected in Chapter 3.
Chapter 6 treats Koguryo/Japanese common morphemes.
Chapter 7 tries to determine the Koguryo home land by identifying words relating to (agri-)cultural items. Conclusion: Liao-hsi + more south.
Chapter 8: Archeology disproves that Korean and/or Japanese are ultimately Altaic, and the linguistic evidence is weak. The proposed Koguryo-Altaic etymologies are either wrong, loans, or language universals. The Koguryo-Japanese etyma "are not only consistently close both phonetically and semantically, they exhibit shared innovations found only in Japanese and Koguryo".
Chapter 9 delves into the nature of convergence (borrowing). The author basically proves that there is no "Altaic".
Chapter 10 has a go at `mixed languages':
1. The examples Thomason and Kaufman (1988) base their theory of creoles on are non-existent or misinterpreted.
2. Swadesh lists are claimed to contain the the most stable words in any language and are used to guide genetic-linguistic research.
     The author shows, however, that the 12 most frequently used words in a language always go back to its genetic ancestor. It is proposed that such high-frequency lists should be used instead of Swadesh lists. [DG: I fail to see what this has to do with mixed languages.]
Chapter 11 concerns major problems with East Asian linguistics.
1. Overly reliance on the HSR (Historic Sinological Reconstruction), the standard phonetic reconstruction of Old Chinese.
2. Reliance on `unique sets of features' to relate languages.
Chapter 12: There is the objection that the place names on which the correspondence Koguryo-Japanese is based, were from an older people that were replaced by the Koguryoans. However, there are a baker's dozen of `real' Ancient Koguryo words preserved in Chinese texts from the 3rd c. AZ, and more than half of them occur in those place names, which is the end of this objection.
     The author points out the symmetry between the east and the west end of Eurasia, comparing the process of the settling of Japan by the Yayoi to that of England by the Germanic tribes.
     A list of about 170 Ancient and Old Koguryo words and morphemes terminates the book.
     The book is peppered with good advice (and stern warnings) on how to show relatedness between languages.
     [DG: On page 225, after having shows that numerals are easily borrowed, the author argues that personal pronouns can also be easily borrowed because they form a numerical sequence: 1st person, 2nd person, 3rd person. etc. I do not think that the naive native language user is aware of this numbering. I think that the paradigm lists as we know them are an artifact of education.]

* Riley, B.E., Aspects of the Genetic Relationship of the Korean and Japanese Languages, (thesis), 2003, pp. 255.
Chapters 1 to 6 cover anthropological evidence, prehistoric contact, archaeological evidence, the history of Korea, historical records of Japan, and methodology.
     Referring regularly to Antilla (1974) (no literature reference given but I suppose Anttila, "Historical and Comparative Linguistics", 1974, is meant), it is argued that the confusion around the evidence for a Korean-Japanese relationship is at least partly caused by internal dialect borrowings, which are much harder to detect than external borrowings.
     Following Thomason and Kaufman (1988) the author argues that both extreme borrowing and extreme language shift can lead to a language that is not related to any of its source languages, i.e. its origin is non-genetic (p. 70).
     The traces of Austronesian found in Japanese are substratal elements from an Austronesian language present in Kyushu when the Japanese arrived. This is supported by the fact that the original population of Kyushu were Sundadonts.
     Chapter 7 is an extensive, 30 page survey of Old Japanese, presenting several argued interpretations of its phonetic system. The internal reconstruction by Whitman (1985) is accepted.
     Chapter 8 is an even more extensive, 40 page survey of Early and Middle Korean. There is less material relating to Old Korean than to Old Japanese, and because the Korean syllable structure is more complicated than that of Japanese, its rendering through Chinese characters is less informative. Several vowel systems and several causes of vowel harmony are discussed; here the author is visibly hindered by not having a Hangul font. The "Hunmin chŏngum" (hun.min.jeong.eum) uses three distinctive features for the description of the vowels, but these are interpreted differently by different researchers. A similar situation pertains to the consonants; usually Ramsey (1993) is followed.
     Chapter 9 reviews the etymologies of Martin (1966), Miller (1971), Mathias (1973) and Whitman (1985). It is striking that much more work is done to make the Japanese forms match the Korean ones than vice versa.
     Chapter 10 presents the author's own hypotheses and etyma, often following Serafin (1996, not publicly available). Many make use of metathesis, for example
Etym. 135. TWO: pKJ *tukwor, from which
1. pre-MK ?*twup(w)ul → MK twul 'two';
2. pre-OJ *tupa → (metathesis, perhaps after OJ pito 'one') → OJ puta 'two'.
Especially the new hypothesis pKJ *kw → MK k ~ OJ p, combined with metathesis is very fruitful, yielding many etymologies of basic words (several body parts), for example
Etym. 97. NOSE: pKJ *kwongo, from which
1. pre-MK *kwongo → (denasalisation of the ng) → MK kwoh (→ (metathesis) NK kho 코 'nose')
2. OJ pana 'nose'.
     Another powerful hypothesis is pKJ *rS → MK r ~ OJ S, where S is one of the stops k, t, and p; this yields f.e.:
Etym. 9. BLOW: PKJ *purk-, from which
1. MK pwul- 'blow';
2. OJ puk- 'blow'.
     A more shaky hypothesis holds that pKJ had a sound K which reflects as t in MK and as k in OJ; yet 17 examples are found (e.g. MK twon 'money' ~ OJ kane,kana 'metal').
     The vowels correspondencies from Serafim are explained, and new etyma using them are shown. The author also adds four new correspondences for MK wo, with 30 etyma in total.
     As to dialectal borrowing, the author shows various ways how sound rules could be seemingly violated. A simple example would be the sound law pKJ *kw → MK k ~ OJ p, which would reflect *kw as p in Koguryo, Paekche, and Kaya, but as k in Silla. Now, if Silla would borrow a word from one of the other states, a word in which pKJ *kw corresponded to p, would end up in MK. Other more complex examples are given.
     A three-page summary of the origin and development of the Japanese language and a list of 146 etyma conclude the thesis.
     [DG: The above etymologies and others in the thesis look very realistic, based as they are on non-trivial sound laws. On the other hand one wonders whether these sound laws are not so powerful that they allow almost anything to be connected. And why is there no follow-up to this?]
     [DG: The text of this thesis suffers from gross editing and bad proofreading, resulting in awkward sentences, which forces the reader (me!) to read many passages twice. This lowers the quality, credibility and accessibility of this work. The university and the supervisor have failed this student.]

* Georg, S., Japanese, the Altaic Theory, and the Limits of Language Classification, in Perspectives on the Origins of the Japanese Language, 2003, pp. 22.
Even before the idea that Turkish and Mongolian were related was put on paper, it was already criticized by the Prussian Peter Simon Pallas in 1776. The notion that Turks and Mongols were "close cousins" originates from the Tatar historian Abu'l-Ghazi Bahadur Khan in 1659 (not based on language, though), and even in 1836 the Altaic languages were called the "Tatar languages". It was the Finn Castrén who first used the term "Altaic" in 1844 (which then included Finn-Ugric) and applied the comparative method to it. The comparative method was much more successful in the Finn-Ugric part, and it was the Finn G.J. Ramstedt who around 1900 applied the comparative method to the Altaic part. This work was brought to great heights by N. Poppe in 1960.
     Remarkably Poppe's work was heavily criticized by several of his colleagues, remarkably because similar comparative works for Uralic, Afro-Asiatic, Austronesian, even Austroasiatic and Niger-Congo were accepted, and resulted in well established families. The critics maintain that the Altaic languages indeed show relationships but that most of the evidence for it can be better explained by large-scale mutual borrowing rather than by a common genetic origin. But even the critics concede that when all borrowing have been removed, genuine cognates may be left.
     The first problem with the Ramstedt-Poppe sound correspondences is that they exhibit "gaps", situations for which no correspondences exist (the rules for Indo-European, Uralic, etc. do not have gaps). For example, there are zero examples of Turkic-Mongolian etymologies starting with m, g, or n. This seems mysterious until one realizes that Turkish does not have words starting with m, g, or n, and assumes that all Turkic-Mongolian etymologies are actually borrowings. Then the mystery dissolves: Turkic could not lend them to Mongolian, because it just did not have them.
[DG: This raises questions. Suppose there was a Turkic-Mongolian proto-language. Then it must have had words starting with m, g, and n, or Mongolian would not have them (unless Mongolian developed them otherwise). One can easily suppose that these initial sounds were dropped in Turkic, especially if the g was actually ng. This would result in Turkic-Mongolian cognates in which Mongolic initial nasals correspond to nothing in Turkic. Do such etymologies exist? The absence of such etymologies is necessary for the above objection over "gaps" to hold.]
     A second problem concerns vowel length. Turkic and Tungusic have two vowel lengths, Mongolian only one. Now Turkic words with a long vowel correspond to words with short vowels in Tungusic, as in
     T. kādın ~ M. qadum ~ Tg. kadum `father in law',
where one would expect Tg. kādum. The simplest explanation is that the Turkic word was borrowed by Mongolian, where its long vowel was shortened, and then was borrowed as is by Tungusic.
     A third problem concerns the distribution of l, r, š, and z over Old Turkish, Chuvash, and Mongolian, the "rhotacism/zetacism" problem. (See Tekin (1979) for details.) The author offers two Chuvash words that are pertinent to this problem: pir `cloth', and erni `week'. The first is a loan from Arabic, bazz, the second a loan from Persian, āðīne. In both cases a z-like sound turns up in Chuvash as r (rhotacism), which suggests that the words are loans and not cognates. (But see Vovin `Fabrication ...' (2018)).
     Attempts at language classification start with a `diastolic' (expanding) phase, in which more and more languages are added to the comparison. For Altaic this meant that it started out including also the Finn-Ugric languages. The diastolic phase is then followed by a `systolic' (contracting) phase, in which proofs are tightened, speculations are abandoned, and languages are removed from the comparison. For Altaic this meant the Core Altaic of Ramstedt-Poppe. Again a diastolic phase followed in which attempts were made to add Korean and Japanese to the mix, as espoused by Sergei Starostin (1991).
     An extreme form of diastolic language classification is Nostratic, named so by Holger Pedersen in 1903, and elaborated by Vladislav Illich-Svitych in 1971. Although Nostratic includes very many languages, it is based on the comparative method and it is not to be confused with the results of mass comparisons. Criticisms of Nostratic include:
1. Nostraticists compare and reconstruct only words, whereas the comparative method gets more power out of morphemes than out of lexemes.
2. The shear amount of work makes it very difficult to avoid errors.
3. Often single language items are taken as representative for a whole family, leading to skewed results.
4. Words are sometimes segmented incorrectly. A good example is Turkic yagmur `rain', the -mur of which is compared to Korean mul `water'. However, Turkic yag- means `to fall (said of water)' and -mur is a substantivizer.
     To the author the "use" of language classification is that it explains facts about concrete, existing words that need scientific explanation (like why we have Eng. that ~ Dutch dat ~ Ger. das/dass). Nostratic and the theory behind it does not seem to provide such explanations. Also, Nostratic uses the comparison of reconstructed proto-forms, which are not concrete, existing words, and which as such have lower status and proving power. Reconstructed forms are more malleable than real-world words, and the temptation exists to bend them to better fit a higher-level comparison.
     To test the quality of the etymologies in the Macro-Altaic of S. Starostin, the author examines 11 etymologies of very stable words that have representatives in at least four of the five language groups. Ten of these are shown to be faulty and are rejected; some common faults are violation of Starostin's own sound laws, and lax semantics: for example `eye' matched to `tears', `one' matched to `each', etc. The one surviving word is `stone', where at least the correspondence Turkic-Mongolian is OK, except that the author shows that the Mongolic form is a direct loan from Chuvash Turkic. For people who cannot believe that a language would borrow a word like `stone': the Romance languages borrowed the Greek word `petra' instead of continuing the Latin `lapis'.
     The author looks forward to developments in the comparison of Korean and Japanese, but even if it turns out that "a given language or family is not demonstrably related to a given other one - or to any other one" (his words), that is "not a nuisance, but a finding."

* Reckel, J., Korea and Manchuria: The Historical Links between Korea and the Ancestors of the Modern Manchus, 2001, pp. 12.
Although Korean tradition speaks of more than 4 millennia of history (from 2333 BZ, when Koreans entered the Liao basin, to today), historical facts and linguistics tell a different story. For one thing, Korea as a unity did not arise until around 1000 AZ, and the Yalu and Tumen rivers were recognized as national borders only around 1400 AZ.
     The earliest Korean kingdoms arose in Manchuria and north-west Korea around 1000 BZ, and consolidated into the Puyo kingdom in the last centuries BZ. In the 1st century BZ the Maglal entered present-day Manchuria, drove the Puyo kingdom south and became the Manchurian kingdom of Parhae. All these kingdoms were multi-ethnic and multi-lingual, as place names correlating with Old Korean, Old Japanese and a at least a third language show.
     The dispelled Puyo kingdom split into Koguryo, Paekche, and Silla (possibly along ethnic lines). Starting in 660 AZ Silla, with help from China, conquered Paekche and Koguryo, which left two kingdoms: Parhae + part of Koguryo from Manchuria to Pyongyang; and Silla in south Korea. In 928 the Mongolian Khitan tribe conquered the Manchurian part of Parhae, and the territory of Koguryo (between Manchuria and Silla) fell to Jurchen (Tungusic) tribes.
     Around the same time the Silla dynasty came to an end and was succeeded by the Koryo dynasty which took Kaesong as its capital. In 1117 the Koryo managed to dispel the Khitan and Jurchen from around the Yalu river, thus creating a land road to China. Expeditions to the east, intended to dispel the Jurchen tribes along the Tumen river failed, in response to which the Koryo built a 1000 li wall, the "Goryeo Jangseong", from Wonsan to the mouth of the Yalu river. Around 1400 the Choson dynasty made the Tumen river its border with China, thus creating Korea as we know it.
     The rest of the paper (pages 3-12) fill in many details.

* Vovin, A., Japanese, Korean and Tungusic--Evidence for Genetic Relationship from Verbal Morphology, 2001, pp. 20.
Seems no longer available, not even an abstract.

* Stachowski, M., Uralistik und Turkologie - geschieden und doch verliebt, (in German: Uralistics and Turcology - Divorced and Yet In Love), 2001, pp. 18.
The author identifies four factors that make comparative Turcology difficult:
1. Although there is ample old material, most of it is written in ad-hoc alphabets in ad-hoc spelling, so it is difficult to see if an apparent sound correspondence is precisely that or a difference in spelling conventions or derives from incomplete understanding of the alphabets. Most Uralic languages are written in well-known scripts are are recorded orally.
2. Almost all Turkic languages lie in a geographical continuum, yielding extensive opportunities for borrowing. Most Uralic languages are geographically disjoint.
3. Most work on the Uralic languages is done by researcher who are native to an Uralic language. Most work on the Turkic languages is done by non-native speakers.
4. There are far fewer Turcologists than Uralists.
In addition Turcologists have to deal with the whole Altaic issue, unlike the Uralists.
     Concerning the Uralo-Altaic relationship, there are three schools of thought: genetic; possibly genetic + prolonged areal contact; and areal contact. Some parallelisms between Uralic and Altaic that may not be accidental are:
1. The s ~ t correspondence present in both families, researched independently in both; example: Finnish sinun ~ Mongol činu `your'.
2. The ordinal suffix: proto-Turkic *-m-ti ~ proto-Uralic *-mt.
3. The Turkic plural suffix -lAr seems to consist of the two Uralic collective suffixes -l and -r. But there are doubts here: Turkic -lAr implies individual plurality.
Seven more parallelisms are presented, although they seem less convincing.
     These parallelisms 1. are not incidental in that they occur in many languages; 2. cannot have arisen through sheer chance; 3. cannot have resulted from recent superficial contacts. So they must either be the result of 4. a genetic relationship, or of 5. old protracted areal contact. The author's bet is, hesitantly, on the latter.
     The paper ends with a plea for an exchange of ideas between Uralists and Turcologists.

* Georg, S., Haupt und Glieder der Altaischen Hypothese: die Körperteilbezeichnungen im Türkischen, Mongolischen und Tungusischen, (Head and Members of the Altaic Hypothesis: The Body Part Terms in Turkish, Mongol and Tungus), 2000, pp. 42.
Rebuttal of Manaster Ramer et al.'s `Body Part Terms as Evidence in Favor of the Altaic Hypothesis' (1998).
     The Altaic hypothesis and its problems are explained in 3 pages in the best?/worst? German academic tradition, using sentences of between 60 and 120 words each.
     Manaster Ramer et al. claim to have eleven excellent Altaic cognates of body parts: `head', `eye', `ear', `nose', `mouth', `tongue', `tooth', `hair', `hart', `hand', and `foot'. For each of these words, its derivation from Proto-Altaic to Turkic, Mongolic, and Tungusic, as given by MR et al. was examined, 33 in all. Of these 31 were rejected in detail, on the following grounds:
1. sound law violations;
2. use of words from a single language, presented as if they represent the whole branch (in particular from Sayan Turkish);
3. incorrect or ignored segmentation of the words;
4. unjustified insertion of segments.
     Two etymologies were deemed acceptable: Mongolian *hü(n)-sün `hair' could well be related to Tungusic *pufie-ke `hair'.
     The paper closes with the cautionary words that "not the quantity of superficially similarly sounding words, but only the quality of their etymologies" will be paramount in deciding the "Altaic question".

* Vovin, A., Pre-Hankul Materials: Koreo-Japonic, and Altaic, 2000, pp. 14.
By going over the material with a fine comb and using the tentative reconstruction of Late Middle Chinese by Miyake (1999), the author identifies or improves 10 lexemes and one morpheme from pre-Hangul Korean: `head', `hair', `tongue', `rain', `moon', `fish', `abalone', `toad', `little', `big', and a dative-locative marker. For most of these a cognate in Old Japanese is found.

* Georg, S., Seefloth, U., Uralo-Eskimo?, 2000, pp. 13.
The noun suffixes of proto-Samoyedic and proto-Yupik show vague similarities. By decomposing them into their three components (the number of objects, the person of the owner, the number of owners), by applying internal reconstruction, and making a number of reasonable assumptions, the authors manage to uncover enough similarity to suggest noun suffix forms for proto-Uralo-Eskimo. The assumptions are carefully delineated and argued.
     One assumption is that some owner components are "reduced" (i.e. their vowel is ə) and others are "full". Most of the reduced suffix components refer to 1st and 2nd person owners, while the full components are used for 3rd person owners, the suggestion being that the latter were introduced in the language in a later stage. On the basis if this distinction the data suggest that -t- is the object plural before reduced components and -i- is the object plural before full components. This agrees with the Eskimo-Aleut data.
     [DG: It is also possible that -t- is just the old plural particle and -i- the newer one and that they developed hand-in-hand with th plural owner particles. Cf. perhaps Hungarian.]

* Vovin, A., Did the Xiong-Nu speak a Yeniseian language?, (in two parts), 2000, pp. 18+6.
All that is known of the language of the Xiong-Nu is a set of about 150 words, some with translation, and a short "couplet", all in Chinese characters, recorded in the early 7th century AZ, describing events that occurred in 329 AZ. The short "couplet" consists of two groups of five characters; its translation is given as: "Army go-out, bok-kok capture" where "bok-kok" is the barbaric title of Liu Yao (a Xiong-Nu chieftain); number and tense are not indicated in the Chinese text. With so few and ambiguous date speculation is rife.
     The obvious way to see if Xiong-Nu is language X is to translate (variations of) the Chinese translation into language X, and see if it matches. For this to work we need a good idea of the pronunciation of Old Chinese.
     In 1922 Ramstedt tried a match with Old Turkic, but the author rejects it on several grounds, among which a too liberal interpretation of the Chinese translation. In 1948 Bazin again went for Old Turkic, but this interpretation is rejected because it is based on incorrect word division. In 1962 Pulleyblank suggested that the language belonged to the Yeniseian family. Pulleyblank also gives several arguments why the language cannot be Altaic, but the author refutes them all.
     The four-word "couplet" is not a good source of information: the first word is *s(l)us-ke `army', which matches the translation but is a general cultural word, not diagnostic of any specific language; and the third word is bok-kok, a title. This leaves us the second and fourth words, both verb forms of unknown tense. On the basis of Starostin's reconstruction of proto-Yeniseian the author creates an approximation of the text in proto-Yeniseian and then uses information from the Kott side of Yeniseian (Kott, Pumpokol, Arin, Assan) to arrive at a more detailed form. The phonetics of the text is reconstructed as

    *suke  t-i-r-ek-ang          bok-kok k-o-t-o-kt-ang,
    armies PV-CM-PERF-go_out-3pp bok-kok PV-?-OBJ-CM-catch-3pp,
where PV = pre-verb, CM = conjugation marker, PERF = perfective, and OBJ = object marker. The translation is then: "Armies have gone out. They will catch [the] bok-kok." The plural of both subjects and the perfect tense of the first verb form are directly visible as Yeniseian particles; the second verb form does not carry a tense indication, which suggest present or future. Phonetic details (t for s) suggest a closer relationship to Pumpokol than to the other Yeniseian languages.

* Janhunen, J., The Convergence and Divergence of Korean and Japanese, 1999, pp. 23.
Summary of the (known) situation:
1. The linear ancestor of the Japanese language was once spoken in the southern and southwestern parts of the Korean Peninsula;
2. The ancestral stages of Japanese are divided into pre-Proto-Japanese (spoken in Korea) and Proto-Japanese (spoken in Japan).
3. The linear ancestor of the Korean language was originally spoken in the southeastern part of the Korean Peninsula.
4. There is no consensus whether Korean and Japanese have a common ancestor or not.
     Languages can change 1. by divergence (changing their own properties), or 2. by convergence (taking properties of neighboring (related or unrelated) languages).
     The author argues that Japanese is the result of a Sinitic-type language that had a long period of convergence with an Altaic-type pre-Proto-Korean on the Korean peninsula or North of it, followed by 1300 years of divergence after the last non-Korean speakers had left for Japan.
     During the convergence period the Sinitic-type language picked up some Altaic structure like suffixes and word order, and pre-Proto-Korean picked up East Asian features like focus and tones. Also words were exchanged. Differences between the two languages were formed during the subsequent divergence period, mainly in the phonetic systems. Many detailed hypotheses are given. F.e. Korean and Japanese both have short and long vowels, but their origins are different: Japanese long vowels originate from double identical vowels, whereas Korean's long vowels are the result of influence by tones.
     Note the use of `Sinitic-type' and `Altaic-type'. There is no commitment that these languages are in fact Sinitic or Altaic.

* Johanson, L., Cognates and Copies in Altaic Verb Derivation, in Language and Literature – Japanese and the Other Altaic Languages, 1999, pp. 13.
The author points out the importance of finding criteria to distinguish between cognates (genetically related material) and copying (borrowings) and emphasizes that extensive borrowing does not result in genetic relationship (sic!). The criteria the author mentions are cached in vague wordings, e.g. "general tendencies with respect to the susceptibility of lexemes to copying". The only concrete statement the author makes is that "the so-called causative morphemes such as -bū-" ... "give positive circumstantial evidence in favour of of genetic connections".

* Johanson, L., Attractiveness and Relatedness: Notes on Turkic Language Contacts, 1999, pp. 8.
The paper starts with a sharp this-side that-side exposition of the Altaic debate. Observing that the determination of genetic relationship between languages depends strongly on the ability to recognize loans (called here "copied code"), and lamenting that the explanation "borrowing" is used often indiscriminately, the author suggests a property of a language feature that will help to determine how easily the feature will be copied. The property is the "attractiveness" of a feature. A feature is "attractive" if it is regular, transparent, and simple. An unattractive feature in language A is a good candidate for being replaced by an attractive feature from language B.
     In general Turkish is a very regular, transparent, and simple language, and indeed it has copied few or zero features from other languages. The author points out that precisely in the few regions where Turkish is irregular it shows correspondences with Mongolian: in locatives, aspect-mode-tense markers, pronouns, and actionality markers close to the stem. Because these features are unattractive and therefore unlikely to be copied, these correspondences can perhaps be better explained as genetically related than as being copied. [DG: The actual formulation by the author is vaguer.]

* Baxter, W., Manaster Ramer, A., Beyond Lumping and Splitting: Probabilistic Issues in Historical Linguistics, 1999, pp. 20.
To extend the time depth of the traditional comparative method, the authors propose the following scheme to determine the probability that two languages A and B are related. A manageable list L is made of concepts that are known to have a low probability of being borrowed, and the most common translations of these concepts in A and B are determined. Using a suitable function F to compare words in A and B for relatedness, the number of words in list that are judged to be related is counted (= N). To compare this number with the number that would be obtained if A and B were unrelated (the null hypothesis) the procedure is repeated a large number of times (say 10000 times) for random permutations one of the sides in the list. This yields a frequency distribution (determined by L and F), in which the position of N can be looked up. This position (the percentile) gives the probability that the null hypothesis holds, i.e. that the result N could have been obtained by coincidence.
     Following a challenge by Hock & Joseph (`Language History, Language Change, and Language Relationship', 1996), the authors set up such a list of 33 words for English versus Hindi, and using equality of the Dolgopolsky classes of the first letters of the words for F, obtain 9 matches. The randomized runs produces a frequency distribution that matched the Poisson density function (as any statistician would expect). Plugging in the numbers they obtained that getting 9 or more matches out of the 33 words on their list using their function had a chance of 0.018, which is below (but not much) the standard 5% percentile (= 0.02).
     The authors consider the function F. If it is made more loose, the number of matches go up, but so does the tail of the Poisson distribution, and vice versa. Neither the existence nor the meaning of a possible optimum are considered.
     It is pointed out that success on this test foes not prove relatedness, only the strong likelihood of it. Conversely, failing this test does not mean proof of unrelatedness, but rather that this method, list L and this function F together do not produce a positive indication of relatedness. [DG: It is worse than that. Success on the test only means that the list L is not random. It says nothing about the nature of the correlation. Possibilities are: 1. the list was doctored; 2. the list was polluted, by borrowings by one language from the other or vice versa, or by both languages from a third source; 3. the two parts of the list have a common descent.]
     [DG: Stephan Georg has pointed out that the use of language classification is that it explains remarkable facts that need explanation. The trouble with the above form of probabilistic linguistics is that it explains nothing, it does not reveal structure. Including etymological knowledge about languages A and B in the function F would make it sharper, confirming and perhaps refining that knowledge.]

* Therien, T.L., Chickens and Eggs: The Undying Issue of the Primacy of r/l or z/š in Altaic Historical Linguistics, 1999, pp. 23.
See Tekin (1979) for a description of the problem.
     The author starts by making a very strong case for the Altaic hypothesis on the basis of the personal pronouns.
     Because rhotacism and zetacism are more or less frequent phenomena, and lambdacism and sigmatism are rare phenomena, the author proposes to separate the problem into one with the usual components and one with the rare components, and leave the latter to a later day. (See Antonov (2011).)
     To decide between the two hypotheses one could try to find loans in surrounding languages from different ages, to so gauge the development in Turkish and Chuvash. But the author cites an example by Menges (1978) in which the Old Turkish word yüz `hundred' is found in (non-Altaic) North-Samoyedic as yur. Now the only contact between North-Samoyedic and Turkic was a long time ago, so this suggests that proto-Turkic had r, and Common Turkic acquired its z through zetacism. However, the evidence is not unambiguous: North-Samoyedic did not have a z, so if proto-Turkic had yüz and the z was original, it is quite possible that the North-Samoyeds interpreted this as yur.
     The author cites many more examples from Menges supporting zetacism, quoting from several Turkic languages, Mongolian, Tungusic, and Hungarian. Strong evidence comes from Turkish word pairs, related words from the same root. An example is köküz ~ kökräk `chest', which suggests a root kökr which received an epenthetic vowel: kökrköküz [DG: a root kökz with kökzäkkökräk is indeed less probable].
     More support for zetacism (z) is found in the fact that Poppe and Ramstedt have suggested acceptable phonetic values for and : a palatalized r (ŕ), and a palatalized l (λ), supported a.o. by loans into Hungarian: Hung. borju `calf' ~ Turk. buzaγi ~ Mong. biraγu. However, the dynamics of the sound shift ŕz are unclear: ŕ regularly turns into ź in the Slavic languages but never into z. Attempts by Hamp (1971, summarized in Hamp (1974)) to remedy this are sharply criticized.
     Supporting rhotacism, which holds that zr and šl occurred in Chuvash, would require finding out how that happened, but the author notices that apparently no such research has been done.
     Some simple objections to Altaic unity are answered: 1. Turkish, Mongolian, and Manchu don't look alike (neither do Russian and English); 2. Altaic words of Stable Word Stock do not match (Indo-European has 5 words for `moon'). But Altaic has a common word for `hand': Mong. gar `hand' ~ Manchu gala `hand' ~ Turk. kariş `span' ~ Korean karak `finger' ~ Jap. kara `spindle' (and Indo-European doesn't).
     The works of Róna-Tas and Doerfer on the subject receive a highly critical but difficult to follow evaluation. The paper closes with the observation that the pro-Altaicists have the best arguments.

* Stachowski, M., Korean-Turkic Studies, 1999, pp. 12.
The paper consists of two parts, "The š ~ l alternation", and "The S ~ t alternation". Especially the first part is relevant to the rhotacism-zetacism discussion.
     When Old Turkic developed a palatalized n (ń), a new rune (𐰤) was created by adding a stroke to the rune for n (𐰭) (actually ŋ). The author noticed that the rune for š (𐱁) is made by adding a similar stroke to that for l (𐰠), and likewise for z (𐰕) and r (𐰺). Because š is associated with the Altaic and z with Altaic , and the stroke seems to be associated with palatalization, the obvious conclusion is that was a palatalized l (ĺ) and was a palatalized r (ŕ). Now ĺ is unlikely to turn into š, but next to ĺ, which is a voiced palatalized lateral approximant, there is a voiceless palatalized lateral fricative form , occurring, f.e., in some dialects of Ostyak, which sounds much more like š. This is a much better match for .
     In Korean we have the problem that the (unwritten) Altaic corresponds to š (written as a Chinese character) in Old Korean but to l in Middle Korean. Ridiculous theories have been proposed to explain this šl behaviour. The author suggests that in Old Korean times the was still , and was approximated by a Chinese š-like character. Between Old Korean and Middle Korean changed into l, and the mystery is solved.
     [DG: This only works if one assumes that Koreanic preserved a sound as unstable as for at least 2000 years and then, just after it was recorded in a Chinese character (± 0-500 AZ), changed it to l. We are not on solid ground here.]

* Vovin, A., Altaic, So Far, 1999, pp. 59.
This is from a time when the author was a supporter of the genetic relationship of the macro-Altaic languages.
     The paper is actually three papers. The first, 11 pages long, is a bibliographical survey of the individual language groups of macro-Altaic; comparative Altaic; and Korean-Japanese comparisons over the years 1965-1999. Much of this material is not available in English, especially material about minority languages, and a knowledge of Russian, Turkish and Japanese is helpful. Almost all linguistic research on Korean is in Korean, work on Tungusic is almost exclusively in Russian. Mongolian material is available in English, Russian, and Chinese, whereas there is ample material in English on Turkic and Old Turkish. The basic book on comparative Altaic is by Starostin, in Russian (1991), but much material is available in English.

In the second part (10 pages) the author sets out to debunk the "myth" that the Altaic languages would hardly have any words for body parts in common. Starting with a fixed set of ten body parts B[1..10] (“nose”, “eye”, “ear”, “tongue”, “tooth”, “hand”, “foot”, “bone”, “blood”, “heart”), the author defines a kind of "poker hand" [DG: my term] for groups of five related languages/families/phyla L[1..5]. This poker hand is constructed as follows. For each body part B[i] the translations in the five languages L[1..5] are found and clusters of cognates between them are determined inside these five translations. The sizes of the clusters form the property of the poker card.
     For example, for the word “nose” and the five languages Old English, Old Church Slavic, Greek, Albanian, and Sanskrit we get nosu, nos', ῥίς, hundë, and na:s-. Here nosu, nos', and na:s- are related and form a cluster of 3, whereas ῥίς and hundë are further unrelated and form two clusters of 1 each; so this "card" is (3,1,1). This card would be worth less than e.g. (4,1) (i.e. 4 related words and 1 outlier) for supporting a relationship between the languages L[1..5]. We shall see that only 7 different "cards" are possible (for 5 languages). The author now constructs "poker hands" for 8 groups of 5 languages;
1. five modern Turkic languages (Turkish, Kazakh, Tuvinian, Yakut, and Chuvash);
2. five modern Altaic languages (Japanese, Korean, Ewenki, Khalkha, and Chuvash);
3. five modern Indo-European languages (English, Russian, New Greek, Albanian, and Hindi);
4. four Old Indo-European languages and Albanian (Old English, Old Church Slavic, Greek, Albanian, and Sanskrit);
5. five modern Austronesian languages (Squliq Atayal, Tsou, Puyuma, Bunun, and Malay);
6. three proto-Austronesian varieties and two modern languages (proto-Atayal, proto-Tsou, Puyuma, Bunun, and proto-Malay);
7. five proto-Afroasiatic languages (Semitic, Egyptian, Cushitic, Chadic, and Omotic);
8. five proto-Altaic languages (Japanese, Korean, Ma.-Tung., Mongolic, and Turkic).
     The results are summarized in the following table, in which, for example, the 2 in row 1. means that there are 2 words in B[1..10] that have related words in 4 out of the 5 modern Turkish languages mentioned above.

cluster sets              (5)  (4,1)  (3,2)  (3,1,1)  (2,2,1)  (2,1,1,1)  (1,1,1,1,1)

1. modern Turkic           4    2      1      3        -        -          -
2. modern Altaic           -    2      -      -        1        7          -
3. modern Indo-European    -    2      1      1        -        3          3
4. Old Indo-European       -    3      1      2        -        3          1
5. modern Austronesian     -    1      -      1        -        4          4
6. proto-Austronesian      1    1      -      -        -        5          3
7. proto-Afroasiatic       -    -      -      3        -        2          5
8. proto-Altaic            1    1      -      3        1        4          -
A number of conclusions can be drawn:
1. The best "hand" is held by a well-establish family (modern Turkish), as expected.
2. The "hands" of the proto-language families are slightly better than the "hands" of the modern language families. This confirms that the method works: proto-languages are shown to have retained more cognates than the modern languages.
3. Proto-Altaic does better than any of the other proto-language families, having one word in B[1..10] that has a cognate in all 5 languages, and having no word without a cognate in at least one other language. This refutes any claim that in Altaic, body parts words with cognates are underrepresented. Myth debunked.
     (The body parts word that has cognates in all 5 Altaic proto-languages is "nose": PJ: *pana, PK: *koh, PMT: *koŋa, PM: *kaŋ-bar, PT: *kaŋ (the *p in *pana derives from pre-proto-Japanese *kw).)
     Given that there are more cognates in the Altaic branches than in the branches of the other families, the author suggests that the Altaic family, contrary to what Starostin (1991) claims, is not very old, younger than Indo-European.

In the third part (16 pages) the author argues that in examining the relationship between languages "morphological affinity must have precedence over lexical affinity simply because morphology is not so prone to borrowing as is vocabulary", so the author proposes 17 proto-Altaic morphemes. Seven of them are weak because they consist of one letter only and two others are weak because they occur in two of the five families only (Japanese and Korean), but some others are very strong (e.g. perfective *-ta-). In particular the transitivity flipper *-gi-, stands out through its peculiar semantics (common Japanese, Korean, and Manchu+Ewenki).
     The17 morphemes presented here are not evenly distributed: 15 occur in Korean, 14 occur in Japanese, 11 in Manchu-Tungusic, 8 in Turkish, and only 4 in Mongolian. And indeed the author describes this work on morphology as "just a first approximation".
     As to determining the internal structure of Altaic, two things are required: better etymological dictionaries, and emphasis on exclusively shared innovations rather than on common shared retentions.
     The last 19 pages are literature references.

* Georg, S., Michalove, P.A., Manaster Ramer, A., Sidwell, P.J., Telling General Linguists About Altaic, 1998, pp. 34.
Critique of the P.R. of the Altaic hypothesis: 1. everybody has their own version; 2. authority figures contradict each other; 3. the opponents are are outspoken and the adherents are indecisive.
     This is followed by critique of the work of two anti-Altaicists, Nichols, and, particularly, Doerfer. Róna-Tas is praised for his balanced view.
     A few points:
• The phonetic value of in the rhotacism/zetacism debate is immaterial [DG: but: 1. the value can help determine if a word can be a borrowing if the value in the other language is known; 2. a sound law r may be more likely than z], given the value of .]
• It is impossible to disprove the Altaic hypothesis: there cannot be a counterexample. The burden of proof is on the pro-Altaicists; the world can then examine the proof and accept or reject it.
     [DG: Georg has since moved to the anti-Altalic camp.]

* Oh Sang-suk, The Korean Vowel Shift Revisited, 1998, pp. 19.
Donegan's Natural Phonology (see Donegan (1978) in NatLang) is a theory that describes and limits how phonemes may change. Remarkably, all changes make the pronunciation easier, which seems paradoxical, because nothing goes downhill all the time. The two most important classes of change are `fortition' and `lenition' [DG: no idea why they are called that]. Fortition simplifies the pronunciation of the phoeme by removing a feature (or perhaps replacing a feature by one requiring less effort): e.g. the [+rounded] is removed from the ø, turning it into an e. It tends to reduce the number of phonemes. Lenition simplifies the pronunciation by spreading a feature over the neighbouring phonemes: e.g. the feature [+fronted] of the i in oi may spread over the o, turning it into an ø. It tends to increase the number of phonemes. French is a simplification of Latin, but still it has many more phonemes.
     The author criticizes Lee's (`Geschichte ..., 1977) vowel shift on several grounds:
1. the shifts are not motivated;
2. an argument for the system was vowel harmony, but there is no proof that Old Korean had vowel harmony;
3. the phonetic values are based on those of Old Chinese and Mongolian, but these are not known with any accuracy; also, Old Chinese underwent a vowel shift itself, as evidenced by cognates in Taiwanese and loans in Vietnamese;
4. Lee bases his shifts on a theory of Martinet, which by itself is unsubstantiated.
Also, the interlocking push chain and drag chain don't work. The author then shows that several of Lee's vowel shifts violate the rules of Natural Phonology, and concludes that there was no vowel shift.
     To explain the vowel harmony in Middle Korean in the absence of a vowel shift, the author proposes a "sonority" opposition system, in which ㅗ, ㆍ, ㅏ resp. have more sonority than ㅜ, ㅡ, ㅓ, which matches the traditional division into bright and dark vowels.

* Gruzdeva, E., Nivkh, Lincom Europa, Munchen, 1998, pp. 66.
Concentrates on two dialects of Nivkh: Amur and East-Sakhalin. They differ about as much as Dutch and (real) Flemish. One problem with this book is that only occasionally does the author indicate from which dialect a given example is taken; this keeps the reader guessing and looking for clues (doable). It turns out Nivkh is a language with simple and systematic morphology, somewhat like Turkish, hence the modest size of the book. One peculiarity is that even endings contain highly marked consonants, for example the ħ; this is cross-linguistically rare. An example is the ending X-roħ `into X', but the riddle is soon solved: as a noun roħ means `inside', and the "endings" are actually compounds, possibly in the process of grammaticalization. Another peculiarity is the large-scale mutation of initial consonants, which is very reminiscent of Irish.

* Ruhlen, M., The Origin of the Na-Dene, 1998, pp. 3.
Actually a comparison of Yeniseian and Na-Dené.
     Provides 36 Yeniseian/Na-Dené cognates, most of good quality, including a rule for the place of the glottal stop (if present) in a stem: before the vowel in proto-Na-Dené, after the vowel in Yeniseian.

* Manaster Ramer, A., Vovin, A., Sidwell, P., On Body Part Terms as Evidence in Favor of the Altaic Hypothesis, 1998, pp. 23.
Disavowed by Vovin.

* Witczak, K.T., Review of Vovin `A Reconstruction of Proto-Ainu', 1998, pp. 11.
Admitting that (almost) any publication on Ainu is an improvement, the reviewer is highly critical of the book:
1. Vovin ignores (or is ignorant of) important material, for example the Ainu publications by Alfred F. Majewicz, Piłsudski's Sakhalin Ainu material of 1912, or the Old Ainu data, collected by Girolamo De Angelis in 1619-1621.
2. Some phonetic analyses are sloppy, for example concerning t- ~ r- ~ tr-.
3. The suggestion of a relationship between Proto-Ainu and Proto-Austronesian is based on totally insufficient evidence.
     To illustrate this last point, the reviewer makes a stronger case for an Ainu-Indo-European relationship, using Vovin's own material. [DG: If these etymologies hold water, they can only be used as support for Nostratic, I'd say from the looks of them.]

* Maher, J., North Kyushu Creole: An Hypothesis Concerning the Multilingual Origins of Japanese, 1998, pp. 22.
Not available. See Maher (2009).

* Unger, J.M., Rejoinder to Roy Miller's review of `Literacy and Script Reform in Occupation Japan', 1998, pp. 2.
[DG: I have kept this as a shining example of the style of discussion among Altaic linguists around the turn of the century. Cf. Robbeets et al. 2019.]

* Martínez, G., Origin Theories of the Korean Language, 1998, pp. 7.
Short overview of four theories:
1. The Koreans came from the east (i.e. Japan). This view is largely colonialist. People in those regions moved from west to east and not the opposite.
2a. The Koreans came from the west (i.e the Altaic mountains). Around the third millennium BZ people living near the Altaic mountains began to migrate eastward, came in contact with Paleoasians, reached the Korean peninsula.
2b. The Koreans came from the west but by a more southern route and their language is related to Dravidian, as shown by typological and lexical correspondences.
3. The Koreans came from the south, from the South Chinese coast, and the language is Austronesian. The evidence is linguistic: open syllables, the honorific system, numerals, and several body parts; and anthropological: rice cultivation, tattooing, and the myth of an egg as the birth place of royalty.
4. The Koreans came from the north, again from the Altaic mountains, and their language belongs to the Altaic family. This is supported by 250 years of research.
     None of these theories is universally accepted but the Altaic theory isthe most accepted.

* Janhunen, J., Problems of Primary Root Structure in pre-Proto-Japanic, 1997, pp. 17.
Observes that many Japanese words are monosyllabic, and even more words seem to be compounds or reduplications of monosyllabic words, where meaning can be attached to the parts: -ma `place', as in yama - `mountain', sima - `island', hama - `beach', numa - `marsh'; or -to - `door' as in ido - `well', mado - `window', yado -`shelter', kado - `gate'; and tens of others.
     The author then hypothesizes that pre-Proto-Japanese was primarily monosyllabic (CVC), lost the final consonant (which in some cases changed to y), and had to resort to reduplication, compounding and introducing tones, to fight the arising homonyms, much like Mandarin Chinese did.

* Manaster Ramer, A., Sidwell, P., The Truth about Strahlenberg’s Classification of the Languages of Northeastern Eurasia, 1997, pp. 22.
The first to suggest/claim that Strahlenberg believed in the genetic unity of the Altaic languages was Donner in 1901, and almost all authors after him copied that notion.
     However, Strahlenberg just recognized 6 essentially unnamed "classes", numbered 1-6. They encompassed the following classes of today: 1. Finno-Ugric; 2. Turkic; 3. Samoyedic; 4. Mongolic (including two Sino-Tibetan languages); 5. Tungusic (including Kott, Arin, Koryak, and Ainu); 6. peoples living in the extreme Northeast (which included several North-Caucasian languages). Some relationships between these groups were suggested by Strahlenberg, f.e. Finno-Ugric and Samoyedic, but as peoples, not linguistically.
     This is substantiated by ample quotes from Strahlenberg's original work.

* Manaster Ramer, A., Nostratic from a Typological Point of View, 1997, pp. 26 (79-104).
Cited here for the "Polish Village" argument on pg. 82; for the full annotation see the entry in the file NatLang.html in this directory.
     An argument used by anti-Altaicists to explain the differences between Chuvash and Common Turkic involves the sound law ŕz (zetacism), but some scholars claim this that sound law is impossible. However: some Polish dialects changed žz, and later most Polish dialects changed ŕž, but about 20 polish villages took these changes in in the reverse order, so they underwent ŕžz, i.e. zetacism. Unfortunately the author does not name any of these villages.

* Vovin, A., Voiceless Velars in Manchu, 1997, pp. 18.
Some confusion surrounds the two voiceless velars in Manchu, /k/ and /x/. Both proto-Tungusic and Manchu have /k/ and /x/, but words in proto-Tungusic that have /x/ have lost the /x/ in Manchu. The Manchu /k/ and /x/ both come from proto-Tungusic /k/, but why did some /k/s turn into /x/ while others stayed /k/ (both in initial and intervocalic position). To confuse matters further, the Manchu /x/ is transliterated as h, suggesting that it is a laryngeal, which it isn't.
     The most striking opposition of /k/ and /x/ is in the perfect participle. Most verbs have a /-xa/, /-xe/, or /-xo/ ending for it (controlled by vowel harmony) but a minority has /-ka/, /-ke/, or /-ko/. The author shows that native stems in Manchu can only end in a vowel or n, and proposes that originally all stems used /-ka/, /-ke/, or /-ko/. In verbs stem ending in a vowel the /k/ became intervocalic and changed to /x/, whereas in stems ending in an -n, the -n prevented the /k/ from becoming intervocalic. At a later time the n disappeared. This is supported by related words in which the n is still present. A very small number of verbs have /-ngka/, /-ngke/, or /-ngko/; these are explained as having ended in -Vne.
     A second peculiarity of the Manchu verb is that many transitive verbs start with /x/ and many intransitive verbs with /k/. The author suggests that in proto-Tungusic, verbs could have an object prefix (a feature areally known from Nivkh and Ainu), which consisted of a single vowel. This vowel caused the initial /k/ of the transitive verbs to be intervocalic, and thus turn into /x/.

* Vovin, A., The Origin of Register in Japanese and the Altaic Theory, 1997, pp. 21.
The Japanese tone system shows itself most prominently in two-syllable words followed by a (on-syllable) particle. Each of the syllables can be High or Low. This allows 2×2×2 = 8 combinations, of which 6 are reconstructed for proto-Japanese. The combinations HL-H and LL-H do not occur; i.e. once the tone is low the particle cannot be high any more (which is basically still true for Modern Japanese). The tone combinations are divided into two classes, called registers: those that start low and those that start high.
     Proto-Altaic is reconstructed with three kinds of stops: voiceless, voiceless aspirated, and voiced. For proto-Altaic words that begin with a stop, the author shows that there is a strong correlation between the stop being voiceless (both kinds) and the corresponding Old Japanese word having High register. The correlation between voiced and Low register is less tight. This makes the register in Japanese related to words in Tungusic, Mongolian, and Turkic, thus supporting the Altaic hypothesis.

* Werner, H., Die ketische Sprache, Wiesbaden, Harrassowitz, Tunguso-Sibirica, ISSN 0946-0349 ; Bd. 3, 1997, pp. 405.
To do.

* Werner, H., Das Jugische (Sym-Ketische), Wiesbaden, Harrassowitz, 1997, Veroffentlichungen der Societas Uralo-Altaica Bd. 50., pp. 314.
To do.

* Miller, R.A., Languages and History: Japanese, Korean, and Altaic, White Orchid Press, 1996, pp. 244.
Highly polemic discussion of the deplorable state of affairs in the research of Altaic, in which all opponents of the author are eloquently and mercilessly put to the sword, in rolling paragraph after paragraph. Very amusing, but in all this expert lambasting the matter at hand --the relationship between Altaic, Korean, and Japanese-- sometimes gets pushed into the background, or worse, delegated to a reference to one of the author's other works.
     Still, sifting the linguistics from the chaff, many interesting etymologies can be uncovered. The most impressive is the solution to the problem why scribes of Old Korean used the Chinese character shih (= death) to mark words which in Middle and Modern Korean end in an l. The author shows that Proto-Altaic tends to be represented by s or sh in peripheral languages, and elsewhere as l. Taking the use of sh by the Old Korean scribes at face value, he then concludes Old Korean had developed sh from Proto-Altaic by one historical route, and Middle Korean had obtained l from by another historical route, which means that Old Korean is not the exact precursor to Middle Korean [DG: just as Latin is not the exact precursor to Italian: Italian preserves duals like le braccia `the arms', which Latin had already lost]. In short, the Old Koreans did say kish for modern kil `road', but the l did not derive from the sh: both derived from Proto-Altaic . The argument is then extended to relate Korean pyeol `star' to Old Japanese posi, Modern Japanese hoshi.
     Among the chapters are "Japanese and Korean in Altaic", "Altaic in Japanese and Korean", and "Borrowings", which shows how knowledge about the cultures can help identify borrowings. The book has two excellent indexes, one general, and one with the words from the various languages, ordered by language.
     [DG: The author explains the German umlaut as a form of vowel harmony: Ger. der Mann, die Männer, but this runs into trouble with des Mannes. How is the e in -er different from the e in -es? Synchronically it isn't, but of course diachronically it once was, -iz vs. -es. So in a historical sense, the German umlaut could be considered a form of vowel harmony, although it would have to be regressive.]

* Solntseva, V.N., Solntsev, M.V., Genitive Case in Altaic Languages and in Some Languages of Southeast Asia, 1996, pp. 8.
The authors observe that the genitive in the Altaic languages is very often expressed by a suffix ending in -n. They also note that the same or a similar suffix in related languages is used also for forming an attributive adjective, a noun, or a subject.
     Citing examples of morphemes that are prefixes in one language and suffixes in another related language from the Slavic and Baltic languages, the authors connect this to the genitive prefix ɂəng in Taioh and Ruc, two Mon-Khmer languages in Vietnam/Laos.
     These morphemes from various languages are compared in detail and many similarities in function are found.
[DG: The Finnish genitive/accusative in -n is not mentioned.]

* Ramsey, S.R., Some Preliminaries to Reconstructing Liquids in Earlier Korean, pp. 14. 1996,
Not available. No abstract available.

* Schönig, C., Bemerkungen zu den `altaischen' Personalpronomina, (in German: Notes on the "Altaic" personal pronouns), 1995, pp. 32.
Taking "Altaic" as a Sprachbund, the author gives a very detailed account of the pronomina in 9 Altaic languages. A summary of the characteristic forms of the 1st, 2nd, and 3rd person singular follows.

	Turkey Turkish	ben/ban-	sen/san-	o/on
	Orkhon Turkish	bän		sän		ol/an-
	S. Sib. Turkish	män		sän		ol/on-
	Yakut		min/miñ-	än/äñ-		kini
	Chuvash		epĕ/man		esĕ/san		văl/un

	Old Mongulian	bi/min/		ti/tin		i/in

	Manchu		bi/min		si/sin		i/in
	Ewenki		bi/min		si/sin		nuŋan
	Nanai		mi/min		si/sin		ńoan
The pronouns of each of the languages, including their plurals, are presented in a table, specifying all their relevant forms, and each form is discussed and suggestions for its derivation and development are given, often in careful terms.
     The 1st and 2nd pronouns are uniformly B-S (Turkic), B-T (Mongolic), and B-S again (Tungusic), but the 3rd pronouns are less regular. Those in Turkic mean `that one', except the one in Yakut, which is of unspecified origin; the Mongolian and Manchu ones mean again `that one', while the other Tungusic 3rd person pronouns may be related to a word for `the other'. It is noteworthy that Old Mongolian has three stems for 1st person singular (and so does Modern Mongolian).
     The author does not comment on the similarity between the pronouns of Turkic, Mongolian, and Tungusic, beyond saying that the use of labials for 1st person and dentals for 2nd person is so widespread that it is inconclusive.

* Vovin, A., Once Again on the Accusative Marker in Old Korean, 1995, pp. 8.
The accusative in Old Korean texts is marked by two different characters, which the author reconstructs as -γïl and -[l]ïl. Based on the dates of occurrences of the characters the author follows Miller in concluding that the second is a more recent version of the first. The -γïl belongs to Old Korean, and given the relation of proto-Korean b/p to Old Korean γ, the form -bïl is reconstructed for proto-Korean. On the basis of Korean dialects, the author divides this in -bï + -l, and suggests that the accusative particle -bï may be related to the proto-Japanese accusative marker *-bo and the proto-Tungusic marker *-ba/*-be. The remaining -l may be an emphasis particle similar to or even related to the Old Japanese -si. So Japanese, Tungusic, and Korean share another important isogloss.

* Hudson, M.J., The Linguistic Prehistory of Japan: Some Archaeological Speculations, 1994, pp. 26.
Theories about the prehistory of the Japanese language can be divided in two main groups: those by Japanese researchers and those by other researchers. Those by Japanese researchers are mainly based on political and nationalistic motives. The others can be divided in three main lines:
1. Japanese is an Altaic language;
2. Japanese is a Southeast Asian or Austronesian language;
3. Japanese is a (mainly Austronesian-Altaic) mixed language.
     The Altaic theory is hampered by the scarcity of convincing cognates between Japanese and the Altaic languages. The Austronesian theory is supported by cultural considerations, e.g. the prototype Japanese house is of Malayan or Polynesian type. Linguistically, Japanese has even fewer cognates with Austronesian than with Altaic, but it shares with Austronesian its phonological simplicity and many other phonological properties. Because these features do not cover the whole language Altaic influences are assumed, often in alternating waves or strata, thus leading to the mixed-language theory. The last big wave was the Yayoi migration, which, according to some researchers, caused such a language upheaval that a "North Kyushu Creole" arose. Solnit (1992) reports that 40% of the Japanese words for which an etymology is available have an Altaic etymology only, 28% Austronesian only, 23% both Altaic and Austronesian, and 9% from elsewhere. [DG: There is considerable overlap here: 63% of Japanese can be explained as Altaic, and 51 % as Austronesian.]
     Japanese and Korean differ far too much to have split off a common source in the Yayoi period (400 BZ to 250 AZ). So either Japanese was born out of the turmoil of the Yayoi migration, or it developed long ago on the continent. A third possibility is the (unrecorded) migration of pre-Japanese to Japan millennia ago, around the time Japanese and Korean split. In that scenario Japanese arrived in Japan not with the Yayoi but millennia earlier [DG: in which case it would have diverged into tens of language groups, and we would not have one uniform Japanese today]. There is archeological evidence of such a migration from comb-patterns and fish hooks from the 4th millennium BZ. At that time the Kikaigashima (known today as Satsuma-Iojima) eruption may have depopulated most of Kyushu and peoples from the mainland may have moved in. The author points out that comb patterns are difficult to identify and have been reinvented often, and are therefore not very diagnostic; there are also other inconsistencies.
     The author objects to the creolization theory; the Yayoi migration was a slow process, lasting about 700 years, with farmers replacing the local population, so there was no intensive contact in which immediate reactions were required. [DG: "You come tomorrow?" "No, no come tomorrow, come Friday."] The Malayan-house argument is also rejected: before Yayoi, pit houses were in fashion.
     The author compares the language situation in Japan with that in New Guinea, another island that was (relatively) isolated for perhaps 10,000 years. In New Guinea we find about 600 languages, in Japan 3: Ainu, Japanese and Ryukyuan. This requires an explanation. [DG: The cases are not really comparable: travel is almost impossible in New Guinea, favouring the development of local languages, whereas in Japan a language of a vigorous village could easily overrun the (perhaps completely different) languages of neighbouring villages, thus eliminating those languages.]
     It is clear that Ainu was a language of the Jomon period, but was it the only one? There were the Emishi, who may or may not have been related to the Ainu. [DG: See Schmidt & Seguchi (2014) for skeletal and DNA research indicating a great variety in the population of Japan in the Jomon period.] Judging from old place names, Ainu was uniformly present on Hokkaido and the north of Honshu; the names begin to thin out below the line Niigata-Fukushima. But other languages may have been spoken in these regions too.
     Glottochronology puts the separation of Japanese and Ryukyu at between 1500 and 2000 years ago, which places it at the latter half of the Yayoi migration. This points to Japanese having arrived with the Yayoi, but where did it come from? It is possible that it derived from the language of Koguryo, or of Puyo, Okcho, Ye, or Maek, languages of which a 3rd century Chinese source reports that they were very similar. However, the mentioned realms were situated in northern Korea or Manchuria, not the right place to take sail to Japan from. Place names from that area have yielded words for four numerals, 3, 5, 7, and 10, that look very similar to the corresponding Japanese numerals, giving support to idea that a Japanese-like language was spoken on the Korean peninsula in the Koguryo era (1st to 7th century AZ).
     The author hopes that mass comparisons as proposed by Ruhlen will shed more light on the structure of Altaic. This would make it easier to determine the positions of Japanese and Korean.

* Vovin, A., Is Japanese Related to Austronesian?, 1994, pp. 32.
The author begins by pointing out that any comparison of Japanese to another phylum must start from proto-Japanese rather than from Old Japanese.The main difference is that Old Japanese *b, *d, and *g derive from proto-Japanese *np, *nt, *nk. Ignoring this leads to faulty reconstructions.
     Benedict's (1990) attempt to show a relationship between Japanese and proto-Austronesian fails because of etymologies that do not follow Benedict's own rules; examples are given. Kawamoto's (1993) attempt fails because his proposal has so many non-predictive rules, that almost anything can be proved. Kawamoto's comparison also suffers from incorrect proto-Japanese reconstructions; examples are given.
     There are lexical ghosts in Benedict's work: words that do not exist in the form cited. In addition to lexical ghosts Kawabata's work contains semantic ghosts, words cited with a widely deviating meaning. An example is a word meaning `fiber' cited with the meaning `hair'.
     After these methodological objections, the author provides proto-Altaic etyma that fit better than proto-Austronesian ones for: personal pronouns, numerals [DG: although these are strained], case markers, verbal markers, and four pages of lexical items.
     The author concedes that there might be a proto-Austronesian substratum in Japanese, but remains sceptical. For several words that look as if they might be substratum words, acceptable Altaic etyma are given.

* Vovin, A., A Reconstruction of Proto-Ainu, 1993, pp. 219.
The book covers the reconstruction of the phonetic system and vocabulary of Proto-Ainu, based on old Ainu material from Japanese sources from the 19th century, and glossaries of the modern Ainu dialects. The reconstructed forms are then used to suggest that Proto-Ainu may be genetically related to Proto-Austronesian.

* Vovin, A., About the Phonetic Value of the Middle Korean Grapheme Δ, Bulletin of the School of Oriental and African Studies, University of London, vol. 56, #2, pp. 247-259. 1993,
Several hypotheses have been offered over time for the phonetic value of the Middle Korean grapheme ㅿ:
• Traditionally it is interpreted as z, the result of voicing of s between vowels or after sonorants, i.e. an allophone of s.
• Based on transcriptions in 17-th century Chinese texts, ž and a retroflex s were proposed.
• On the basis of MK moㅿom `heart', `soul' ~ Jap. mune `chest' Starostin suggested = ñ.
The last two are external evidence, but the author adduces Korean-internal evidence for = ñ, which should be better.
     The author collected all 79 words from the 15-th century with ㅿ, representing ㅿ in advance as ñ, which is very confusing. A simple comparison between MK, Seoul, Kyengsang, and Hamkyeng shows that corresponds to s or nothing. However, -ㅿG- (with G a notation for "probably γ") sometimes shows up as ŋ, which gives credence to = ñ. [DG: The rest of the internal evidence is wafer-thin, and can often be better explained by assuming that is related to s.]
     More evidence can be obtained from a Chinese description of Korean sounds, where the is said to have the begin sound of 日 nil (today ri) `sun'. More circumstantial evidence is given.
     See also Robbeets (2005, `Is Japanese ...'), pg 60-62.
     [DG: Hard to believe. The evidence of a relationship between ㅿ and s is overwhelming, and the shape of the ㅿ (ㅅ with bar below) matches those of ㅅ (ㅅ without bar) and ㅈ (ㅅ with bar above).
     Typologically ð (a voiced dental fricative) is the obvious choice, filling the gap between β (a voiced bilabial fricative) and γ (a voiced belar fricative). The `Hunmin jeong-eum haerye' describes the ㅿ as a "half front-tooth sound", which sounds applicable to ð. And the ð definitely has a relationship to s and z. I am amazed that nobody has proposed this yet. Also nobody mentions that the Korean s is not a real /s/, and is more approximant than sibilant, so why should its voiced form be a simple /z/?]

* Doerfer, G., The Older Mongolian Layer in Ancient Turkic, 1993, pp. 8.
The author sets out to determine from which Mongolian languages/dialects the Turks borrowed words when they lived in the Orkhon valley in central Mongolia from the 6th to the 11th century, and finds that the loans were mostly from Tabγač, Qitai [DG: I suppose this is the Mongolian language spoken in the desert county of Qitai], and Yüan Mongolian, the precursor to the language of Genghis Khan.

* Kortlandt, F., The Origin of the Japanese and Korean Accent Systems, 1992, pp. 8.
There are three main types of Japanese pitch profiles: Kyoto-type, Tokyo-type, and Kagoshima-type (which includes Ryukyu). Kyoto-type splits in two subtypes: Kyoto proper and Kochi; Tokyo-type splits in Tokyo proper and Aomori, but in both cases the differences are minor. All 5 types derive from the Old Kyoto pitch system. The original Korean pitch system consisted simply of high and low syllables.
     To compare both pitch systems, 66 pairs of Korean-Japanese look-alikes (for example MK ënyi/ J ane ‘older sister’ and MK kuy `ear' / J kiku `to hear') are compared for pitch. For 58 of them the pitches are similar, for the other 8 they disagree.
     The last paragraph is about the origin of the clearly related pitch systems, but is vague. The author suggests a relation with the glottalized (creaky voice) and aspirated (breathy voice) vowels in some Altaic languages and with the consonant structure of the morphemes. [DG: Aspirated vowels occur phonemically e.g. in Dukha (Turkic) (see de Mol-van Valen, 2017), in Gujarati and White Hmong. The Wikipedia mentions Jalapa Mazatec as a language with phonemic creaky voice; it occurs non-phonemically in Korean after tense consonants with some speakers.]

* Ščerbak, A.M., Sur le soi-disant rhotacisme mongolien, (in French: The So-Called Mongolian Rhotacism), 1992, pp. 4.
[DG: This is not related to the rhotacism/zetacism debate, but pertains to Mongolian itself.]
     The author points out that, in the category words with Turkic z and Chuvash r, there are numerous examples of Turkic-Chuvash pairs, perhaps about 30 Turkic-Mongol pairs, and perhaps a dozen Mongol-Tungusic pairs. This is because the Mongolian and Tungusic words are loans.
     Both s/zr and rs/z are possible, but rs/z (in a few French words) is much rarer than s/zr (general, in Latin and in German). The s/zr is explained as "successive voicing of the s/z under special conditions of consonant weakening" [DG: I don't think this gets you any further than a ř].
     B.Ya. Vladimirtsov (1929) claims that rhotacism occurs in Mongolian, and adduces 8 word pairs as proof, one with r and one with s. One is a clear loan from Turkish; a second is semantically doubtful: mör `track' vs. moski `to track' [DG: This was `voie' and `suivre la trace' in French; suddenly, translated into English, it is not so bad semantically any more!]. Three pairs were rejected because the r did not change to s but was deleted following the s, which came from a suffix. The three others were difficult to explain (away); the author hypothesizes that they originally contained -rs-. It may help that it is known that -ls- reduces to -s- in Mongolian.
     The conclusion is that the existence of rhotacism is far from certain in Mongolian, but that some examples need to be investigated further.

* Martin, S.E., Morphological Clues to the Relationship of Japanese and Korean, 1991, pp. 27.
I hate papers with paragraphs longer than a page, and with 40% of the text footnotes.
     After cautioning that morphological elements are short and their semantics is difficult to assess, and pointing out that morphological elements should not be compared until the earliest vocabulary has been reconstructed, the author sets out to analyze morphological elements, more in particular the basic case endings.
     Taking the case endings of Korean and Japanese in parallel, the author suggests dozens of hypotheses for their origin and development.
     The origin of the recent (17th century) Korean nominative suffix -ka is investigated but no conclusion can be drawn. It cannot have been introduced by the Hideyoshi invasion (1592-1598) because they pronounced their particle ŋa). [DG: Still the date coinides. A hypothesis is that it was originally part of colloquial speech only and was not put in writing before. The Japanese particle ka is linked to the question particle ka [DG: without clear argumentation].
     The Korean object particle is sometimes written hul in Chinese. This may be an attempt to write -ol for which there is no Sino-Korean character, but it is hypothesised that in some circumstances the particle was indeed -hul, and that this is how some nouns that end in -h in Modern Korean but not in Middle Korean got their -h. Drawing on the parallel between -un 1. focus of nouns 2. past participle of verbs, and -ul 1. object of nouns 2. future participle of verbs, the author hypothesises that the adnominal forms may actually have been participles of the auxiliary verb ho-.
     Korean had another focus particle, -n pa (from pa `situation'); this corresponds to Japanese wafapa.
     Furthermore the author makes two noteworthy observations:
• In a sense, all clauses of Korean and Japanese are subordinate; the traditional "sentence" corresponds to the higher-level unit of "discourse".
• The "case particles" of Korean and Japanese are fundamentally different from the case endings of Indo-European. They correspond to prepositions in English, and have the same fluidity in meaning and usage: on vs. at, to vs. for, etc. So it is understandable that Japanese no and ga can sometimes be exchanged.

* Ramsey, S.R., Proto-Korean and the Origin of Korean Accent, 1991, pp. 24.
Middle Korean had an accent system, but today phonemic accents are only found in the North-East (Hamgyeong). The Hamgyeong system is very similar to that of Tokyo:
1 syll.: L-H, H-L;
2 syll.: LH-H, HL-L, LH-L;
3 syll.: LHH-H, HLL-L, LHL-L, LHH-L;
where the last tone belongs to the particle following the word. These patterns can be coded by a single marker on the last H syllable, and the rule that if the marker is not on the first syllable, the Hs start on the second syllable (= if the first syllable does not carry the last H, then the second syllable carries the first H).
     The middle Korean system was more complicated: it had three tones: low, high, and rising, the latter occurring on a long vowel originating from two coalesced syllables (low-high); and it does not allow a simple characterization. The locus of the accent was on the high syllable; what followed was non-distinctive.
1 syll.: L, H, R;
2 syll.: LL, HH/HL-H, LH, RH/RL-H;
3 syll.: LLL, HLH, LHH/LHL-H, LLH, RLH;
     The agreement of tone patterns between Middle Korean and the modern dialects is good for words with an L on the first syllable, but poor for words with H on the first syllable. Looking at words with L or H on the last syllable of the word we find another anomaly: 78% of the one-syllable words and 68% of the two-syllable words have H on the last syllable. The author hypothesizes that these majority words represent the original situation and that proto-Korean had no tone accent but that the last syllable of a word had automatic prominence. The author then identifies influences that may have caused deviations from this original pattern: consonants following the vowel, especially ŋ; reduction from compound words; and derivation/influence from Chinese.
     The Middle Korean verb comes in 8 classes, distinguished by accent patterns, lenition of final consonants, and reaction to the minimal vowel ʌ/ɨ that starts many of the verb endings. The distribution of verbs over these classes is again highly skewed. The author tries to bring order to this chaos by hypothesizing that 1. the proto-Korean verb stem was (almost) always disyllablic, which led to Middle Korean monosyllabic stems through syncope and apocope; 2. the vowel of the second syllable was often the minimal vowel; 3. verbs in which both stem vowels were the same reacted specially; 4. the lenition pairs p/β and t/l originated from *b and *d, resp.; 5. verb stems that start with an aspirated consonant derived from proto-Korean stems that had a suffix from hʌ- `to do'.
     This results in a reconstruction of proto-Korean in which present-day Seoul ssɨ-ta `to write' derives from Middle Korean psɨ-ta H-H, which comes from proto-Korean pɨsɨ-ta LH-H. The author wonders if the minimal vowel ʌ/ɨ derives ultimately from "a fuller set of vocalic distinctions" or it was epenthetic. [DG: It certainly looks that way in modern Korean.]

* King, J.R.P., A Soviet Korean Grammar from 1930, 1991, pp. 26.
Analysis of a 104 pages Korean reference grammar, `Ko.lye Muncen', issued in Khabarovsk in 1930 for Korean Soviet citizens, written by O Changhwan.
     The foreword to the grammar O explains that at that time the spelling of Korean was not fixed and that every writer and teacher had their own rules. This had nothing to do with sloppiness but was caused by the lack of a standard. O also pointed out the unifying importance of standard grammar in a cultural environment. (In Korea itself the spelling was not fixed until 1933.)
     Some points:
• `Hangul' is called Kwukca 국자 `National letters'.
• A principle was that word bases should be written the same under all circumstances, following etymological leads: the word base must be kept separate from suffixes. Examples are: o-as-da `somebody came' rather than was-da (today wass-da), mud-hi-eos-da `somebody was buried' rather than mus-cheos-da, dah-ji-da `sombody was hurt' for today's da-chi-da, etc. Likewise the verbal bases of the irregular verbs are kept constant, regardless of their pronunciation: go-mab-eos-da `somebody was thankful' rather than go-ma-wos-da. The author attributes this to "a confusion about how to write the irregular verbs", but it seems to me that O was at least consistent here, more so than the present official spelling. But the consistency occasionally borders on excess: from ithul `2 days', sahul `3 days', and nahul `4 days' he arrives at onhul `today' (with semantic skew).
• O called the tones of Middle Korean "Chinese character poisoning", and dropped the recording of them. Vowel length indication was rejected because it was not uniform among the people.
• Although O amply uses Hanja in his text, his aim is "to curtail the use of Chinese characters", and he deplores the use of Hanja for native Korean words.
• The "Part of Speech" section covers quite some ground, listing honorific verbs, making dialect comparisons, distinguishing subsentence endings from sentence endings, etc. Some presented endings are very old-fashioned; a strikingly characteristic example sentence is: Ji-geum eun hyeog-myeong si-dae lo-da. `Now is the time for the revolution.' (spacing as in the original), with the archaic ending -lo-da.
• The modern o in the ending -lo is u: Geu-neun u-li chon-eu-lu o-li-la. `He will come to our village.' (with futurum in -li-).
• The "Syntax" section discusses double-subject sentences, pointing out (correctly) that the first subject belongs to the entire sentence and the second subject belongs to the verb only, certainly a deep insight for that time.
     The author points out many Hamkyeng-isms in the text.

* Benedict, P.K., Japanese/Austro-Tai, Karoma, 1990, pp. 276.
Reexamining the Austro-Tai etyma from his `Austro-Thai Language and Culture' (1975), the author adds etyma and extends them to Japanese, postulating an Austro-Tai/Japanese proto-language. The first to split off are the Miao-Yao languages, followed by the Kadai languages, and the third and last split is between proto-Austronesian and proto-Japanese-Ryukyuan. For each cited form it is specified to which of the three levels it belongs.
     Etymologies are based on the comparative method extended with word-restructuring mechanisms, which are very important in Austro-Tai/Japanese. These are: reduplication of the 2nd syllable; reduction of the right, left, or center syllable; and disyllablic drift, in which a monosyllabic word can receive an epenthetic vowel to become disyllablic. The form of the etymologies is as follows: first the reconstructed proto-Austro-Tai form is shown, with information on how it is based on words from various branches of Austro-Tai, and next the sound rules of Chapters 6 to 9 are used to derive a Japanese word. Which rules are used exactly and in which order, is, however, hardly ever made explicit. Representative etymologies are:
STAR: proto-Austro-Japanese buxis → Japanese Fosi
FISH/SQUID: proto-Austro-Japanese śikan `fish' → Japanese *yikaika `squid'
COOK/ROAST: proto-Austro-Japanese talak `cook' → Japanese yak-i `roast'
     [DG: These mechanisms together are so powerful that it seems feasible to derive almost any word from almost any other word. To verify the ± 180 etyma in the Glossary one would need to check every one of them to see that these mechanisms have been applied correctly. A case in point is the derivation of FOUR: proto-Austro-Kadai *śəpat → Japanese , presumably by reduction of left syllable (śəpatpat), loss of final t and aɔ (pat), but there seems to be no rule initial py.]
     Chapters 6 and 7 give the sound laws for vowels and consonants, respectively. The plosives can have an often optional "nasal extension" in many Austro-Tai languages: (m)b, (n)t, (ŋ)k, etc., and there are special sound rules for these. These rules are all rather obvious, and it is only through the word-restructuring mechanisms that many etymologies lead to non-obvious correspondences. Some sound rules are non-predictive: medial q goes to either k or g.
     Chapter 8 supplies details about the "suprasegmental" rules, the three forms of reduction, and how they relate to accent and de-stressed syllables. The Austronesian languages come with three seemingly unrelated accent/tone systems and none of them tallies really well with the Japanese pitch system. The author offers several hypotheses, but in the end only more data will help.
     In Chapter 9 the author tries to find common morphemes in Austro-Tai, but it all is rather speculative; for example the Austronesian actor-focus infix -um- may possibly related to the m of Japanese mi- `to see', as follows: starting from proto-Austronesian kitaɂ `to see', we get with the infix k-um-itaɂ. With reduction of left syllable this yields m-itaɂ, and with loss of the last syllable we obtain mi-.
     In Chapter 10 the author considers the lexicon, and points out that the set of etyma in the Glossary contain almost all major body parts. This is seen as very strong evidence for the validity of the relationship. The discussion of numerals in this chapter is unconvincing. There is also a long discussion about kinship terms.
     Chapter 11 concerns the nature of proto-Austro-Japanese versus proto-Austro-Tai. Japanese seems more akin to Austronesian than to Kadai and even less to the Miao-Yao languages. This may be due the order in which these groups split off, as suggested in Chapter 1, but it may also be an artifact from the lesser availability of Kadai and Miao-Yao data. Japanese seems most strongly akin to the Formosan languages, but the author attributes this to contact during the stay of some migrating Japanese groups in the island of Formosa, rather than to a closer genetic relationship. The Austro-Tai languages have verb-object word order (as has Chinese). The author is vague about how Japanese acquired its Altaic structure and word order.
     Chapter 12 speculates about the trek the Austro-Tai people made, and how a part of them reached Japan. The Austro-Tai people lived somewhere in South-East Asia around 7000 BZ. By 5000 ± 1000 BZ had moved to the Lower Yangtze Basin, having left the Miao-Yao and Kadai behind, and had become the Austro-Japanese. There they became seafarers and groups left for the islands: Formosa, the Ryukyus, the Philippines, Indonesia, and further. Those that stayed behind became more and more Japanese, learned wet rice cultivation, and by 1000 BZ they started to take the rice cultivation with them to Japan, often by way of Formosa, where they exchanged words with the Austronesian-Formosan speaking peoples. This migration gathered speed around 300 BZ, when these Austro-Japanese began to populate Japan in earnest, becoming the Yayoi.
     The stay in the Lower Yangtze basin fits in well with a theory by Ballard `Tonometamorphogenesis' (1983), which explains the exceptional tone sandhi in the Wu language spoken in the Yangtze basin through a substrate from a language with pitch accent. Proto-Austro-Japanese could well have been that language.

* Park Sayhyon, Vowel Harmony In Korean, 1990, pp. 31.
After acknowledging that the mechanism behind vowel harmony in Modern Korean is difficult to explain, the author promises a description that "is simpler and more natural than previous ones". That this simpler description is 31 densely written pages long may be cause for concern.
     Vowel harmony in Middle Korean was based on two sets of vowels, light ones {a ㅏ, o ㅗ, ʌ ㆍ, i ㅣ} (light) and dark ones {ə ㅓ, ɨ ㅡ, u ㅜ, i ㅣ}, and two opposition pairs: a/ə (ㅏ/ㅓ) and ʌ/ɨ (ㆍ/ㅡ). Note that the opposing vowels in the opposition pairs are from different sets.
     The full sets apply stem-internal: all vowels in a stem must either be dark (malssʌm `a saying') or light (təuk `very'). The first opposition pair occurs in the -a/-ə suffix (the INF suffix or extended stem suffix). Examples are tor-a `to turn-INF' versus pur-ə `to blow-INF'. The second opposition pair occurs in suffixes containing ʌ/ɨ (the padding suffix in verbs, and the TOP, ACC, and other suffixes on nouns, e.g. -[n]ʌn/-[n]ɨn `TOP'). An example is na-nʌn `I-TOP' versus nə-nɨn `you-TOP'. This system is simple enough, except that the sets and the opposition pairs do not form natural phonetic groupings.
     In Modern Korean this system has partially collapsed. The ʌ ㆍ has disappeared and was replaced most often by ɨ ㅡ, but also by other vowels. Diphthongs, ay ㅐ, əy ㅔ, oy ㅚ, and uy ㅟ, formed and later reduced to monophthongs, ε ㅐ, e ㅔ, ö ㅚ, and ü ㅟ. And in suffixes ʌ/ɨ coincided into ɨ. So most vowel harmony in suffixes disappeared; only the a/ə distinction in the -a/-ə verb suffix remained.
     In stems the vowel harmony has largely disappeared, certainly in words that contained the lost ʌ ㆍ (Middle Korean malssʌm `a saying' is now malssɨm), but it remained in modified form in sound-symbolic words.
     For sound symbolism the dark and light vowel sets have been extended with the new monophthongs that derive from earlier diphthongs: light {a ㅏ, o ㅗ, æ ㅐ, ö ㅚ} and dark {ə ㅓ, ɨ ㅡ, u ㅜ, e ㅔ, ü ㅟ, i ㅣ}. Examples are ccækkak `squeaking lightly' versus ccekkək `squeaking heavily'; and ppakkom `opening bright small eyes' versus ppəkkum `opening dull big eyes'. (Examples and translation by the author.) [DG: Form and meaning of sound-symbolic words in Korean have a personal component.)
     Next the author discusses five theories presented in the literature and shows that they all fail to characterize the vowel harmony sets adequately.
     The author observes that there is a tendency to raise vowels in Korean; examples are Middle Korean namo `tree' → Modern namu, and maʌm `mind' → maɨm. Next an attempt is made to turn back this raising, in order to reach an older, perhaps more harmony-friendly vowel set [DG: if I understand it correctly]. Many subtle reasonings lead the author to suggest a rather irregular system for Late Middle Korean, with four levels of [height] and four [+back] [+rounded] and three [-back] [-round] vowels, where the [-back] [-round] vowels have an additional feature, [front]. (The reconstructed Middle Korean vowel systems by Lee Soong-Nyong (1955), Moon Yang-Soo (1974), and Lee Kim-Moon (1982) are emphatically and in the case of Lee Kim-Moon vehemently rejected.) More subtle arguments with many examples then lead to the present vowel system and vowel harmony rules.
     [DG: There are so many rules and exceptions in this that I cannot tell if there is a Korean Sound Shift hidden somewhere.]

* Vovin, A., Some Data on the Soviet Korean Language, 1989, pp. 19.
About 150 words, 24 verb endings, and about 90 short sentences of Soviet Korean, collected in Leningrad from direct descendants of parents forcibly transmigrated on Stalin's orders in 1937. The lists are followed by 6 pages of analysis.
     The language is fairly close to the dialect of Hamgyong-bukto (presently North Korea), but shows some Russian influences, for example the loss of the copula.

* Rickmeyer, J., Japanisch und der altaische Sprachtyp, (in German: Japanese and the Altaic Language Type), 1989, pp. 20.
Regardless of whether Turkic, Mongolic and Tungusic form a genetic unit or not, structurally they are similar enough to define an "Altaic Language Type", and the author examines various features of the Altaic languages, comparing them to those of Japanese.
     Phonetics: Japanese and Altaic have very similar phoneme sets; Altaic has vowel harmony, Old Japanese has only traces of it; Altaic has closed syllables, Japanese does not.
Morphology: Structurally they are very similar, but Japanese has prefixes, unlike Altaic, and lacks person endings on the verb, unlike Altaic. [DG: only Turkic!]
Syntax: Altaic and Japanese are both SOV (except north Tungusic); adjectives are closely related to verbs in Altaic, but have separate conjugations in Japanese. Remarkably, negation is expressed differently in the Altaic languages among themselves and in Japanese.
[DG: No mention of relative clauses.]
     All in all, Japanese's similarities to the Altaic type far exceed the differences. Whether this is the result of an old genetic relationship, or characterizes a general language type (shared by e.g. Dravidan, Quechua, Navaho [DG: huh?]) is unclear, but what is clear is that Japanese is not so unique as some nationalists would like it to be.

* Tekin, Talât, Zetacism and Sigmatism: Main Pillars of the Altaic Theory, Central Asiatic Journal, vol. 30, #1/2, 1986, pp. 141-160.
The author sums up reasons why the r is original in the rhotacism-zetacism debate:
1. If the z was original it would have to have changed into r independently in Chuvash, Mongolian, and Manchu, which is absurd [DG: or only in one of them (probably Chuvash) and the others are loans].
2. The same goes for l/š.
3. Old Bulghar loans in Hungarian do not have z/š, but r/l.
4. Bulghar proper names and titles in Byzantine sources have r, not z, i.e. there never was a z in BUlghar/Chuvash.
5. Although in most positions z in Old Turkish, it remains intact in others, mainly when in a consonant cluster, for example köküz ~ kökräk (and several others).
6. Even Doerfer accepts it, finally.
     The words with r in Mongolian are sometimes explained as loans from Chuvash/Bulghar, but this is impossible for geographical reasons: in the middle of the first millennium AZ the Bulghar lived along the Volga, and the Mongolians lived in Mongolia, with at least 2000 km between them.
     Although Doerfer has successfully refuted some Turkic-Mongolic etymologies from the literature, the author claims that some other refutations are faulty, and proceeds to give a lengthy example. Referring to Ligeti, who showed that the j in Hungarian borjú /boryū/ `calf' is of Hungarian origin and does not come from Chuvash, the author objects to Doerfer's identification of as ŕ [DG: but Doerfer got this from the internal proto-Turkic consonant set].
     Next, Doerfer 's etymologies (from 1984) are criticized and then Doerfer's critique of Tekin's etymologies (from 1979) is criticized, in detail, under ample quoting of Doerfer's objectionable text.

* Whitman, J., The Phonological Basis for the Comparison of Japanese and Korean, (thesis), Harvard University, 1985, pp. 251.
Not available. See Whitman (2020).

* Akiba-Reynolds, K., Internal Reconstruction in pre-Japanese Syntax, in Historical Syntax, 1984, pp. 23.
Old Japanese has a case particle ni, the copula ni, and the perfect-tense auxiliary particle nu. The case particle ni is remarkable 1. because of its semantic scope, which encompasses all non-subject cases; 2. because it can be followed by the conjunctive particle te, which normally follows verb forms only.
     To explain these facts, the author proposes a transitive verb *nu in pre-OJ, meaning `to be at' (the location is the object of the verb), and shows that ni, ni, and nu can be interpreted as conjugated forms of this verb. And if ni is a verb form it can of course be folowed by te. [DG: see also Martin (1990) `Morphological Clues ...', pg. 489]
     This removes some syntactic structure from pre-OJ. Furthermore the object marker wo was originally probably an emphasis marker; this leaves the associative particles no and ga as the only particles applicable to nouns. There are two conjunctive particles, te for same-subject conjunction, and ba for different-subject conjunction, but conjunction by juxtaposition was also frequent.
     This paucity of syntactic tools is characteristic of creole languages, and the author proposes that pre-OJ was a creole languages based largely on Altaic (for some of its the structure) and Malayo-Polynesian (for some of its the vocabulary).
     [DG: This is elaborated further by Maher (2009).]

* Doerfer, G., The Problem of Rhotacism/Zetacism, 1984, pp. 7.
For the problem see Tekin (1979).
You wish this paper was about rhotacism and zetacism; it should by rights be called "The problem of Tekin".
     The author examines about a dozen claims that Tekin made in his three papers about rhotacism and zetacism, and casts doubts on them, suggests alternative interpretations for them, or refutes them. This includes claims in which Tekin claims that Doerfer claimed something.
     It is a clear response to Tekin's dogmatism, but that does not mean that the author may not very well have been right, especially when suggesting alternative interpretations.

* Clippinger, M.E., Korean and Dravidian: Lexical Evidence for an Old Theory, Korean Studies, University of Hawai'i, #8, pp. 1-57. 1984,
The paper consists of a 10-page introduction, followed by 3 pages of sound rules and 44 pages of etymologies.
     The author is not explicit about the nature of the relationship between Korean and Dravidian, but suggests (without naming a source) that the Dravidians were not autochthonous to India but came originally from somewhere by the Caspian Sea (implicitly suggesting that they may have met the Korean on their trek to India).
     The similarity is illustrated by a list of about 55 Middle Korean / Proto-Dravidian look-alikes, of varying quality, and by a discussion the typological similarities between Korean and Dravidian.
     There are about 80 sound rules, many with conditions, and all with at least one example. The list of etymologies contains about 400 items, graded in quality classes I to IV. Each etymology consists of the English translation, the grade, the Middle Korean word, corresponding Dravidian words, and a short discussion. Examples are:
• Acquire. I. pat- `to get', Kannada pad.u `to get', ...
• Bites. II. mul- `bites', Tamil mel- `to hold in the mouth', Toda melk `mouthful'.
• Mountain. III. moy, MK mo'lo, Tamil malai `hill, mountain', ...
Etymologies consider the first CVC- only, because they are the most significant parts in both languages.

* Janhunen, J., Kho Songmoo, Is Korean Related to Tungusic?, 1982, pp. 12.
The authors start by pointing out that currently the leading hypothesis is that Korean belongs to the Altaic languages whereas the leading hypothesis should be the null hypothesis that Korean developed independently for longer than the life time of [recoverable] language families. The "proof" for the Korean-Altaic hypothesis rests on "omnicomparativism", which ignores chronological and areal realities.
     The first step in investigating the hypothesis is to compare Korean to its closest neighbour. There is a wealth of comparisons between Korean and Japanese, but these are highly criticized for the artificiality of their complicated phonological explanations. For geographical reasons Kim Dongso (1981) regarded Tungusic a good candidate for comparison, and achieved 183 etymologies between Tungusic and Korean. However, Kim Dongso's work does not recognize loans from third languages, in particular Mongolic and Chinese, leading to anachronistically declaring, e.g. tampae `tobacco' as part of the proto-Korean-Tungusic vocabulary. Two further problems are that the reconstructions of proto-Korean and proto-Tungusic words are skewed to increase similarity, and that the semantics of the corresponding words is very often unconvincing. What remains is a small number of parallels without any phonetic correspondence rules.
     To clarify the status of the Korean language, the following steps would help:
1. The best possible reconstruction of proto-Korean should be prepared from all synchronic, diachronic and dialectal information.
2. Loans from Manchu and Mongolian should be analysed and time and region of contact should be determined.
3. The way Chinese words entered the Korean language should be studied, as far as our knowledge of Old and Middle Chinese dialects allows.
4. The few non-loan Korean-Japanese etymologies with proper matching semantics should not be overlooked (the authors quote here, among others, Kor. cip ~ Jap. ie `house' as such a pair, which seems a stretch).
5. The linguistic connection with Gilyak should not be neglected.
6. In their history Koreans may have come into contact with other groups in Eastern Asia: e.g. Turkic, Iranian, etc.
7. The Koreans arrived in Korea perhaps 2000 years ago. Who did they find there? Korean would have a substrate of their language.

* Pritsak, O., The Hunnic Language of the Attila Clan, 1982, pp. 49.
Thirty-three names of persons are extracted from a large number of sources and a detailed explanation of each name is given. Most explanations involve fragments of Old Turkic and Written Mongolian, and indeed the author determines that the language of the Huns was something between Turkic and Mongolian, most closely to Bulghar/Chuvash. The exact status of this language (Mongolic-influenced Turkic, Turkic-influenced Mongolic, separate branch of Altaic, etc.) is not specified.
     An astonishing amount of information is squeezed from these 33 names. Seven vowels are reconstructed: a, o, u (non-palatal), and e, ö, ü (palatal), with a, e, o, and ö also long. The velars q and G appear next to non-palatals, k and g go with palatal vowels. In endings only i, a, and ā appeared. No diphthongs were found. Over the roughly 200 years of the Hunnic realm diachronic changes can be observed, and many rules are given. A list of perhaps 50 stems is given, but it is left to the reader to determine their meanings from the preceding discussion.
     The author declares Doerfer's paper `Zur Sprache der Hunnen' (1973) to be "disappointing". It seems the feeling is mutual.

* Janhunen, J., Korean Vowel System in North Asian Perspective, 1981, pp. 13.
From a reprint in `Transeurasian Linguistics', Vol. 2.
     Modern Korean has 10 vowels, 5 front ones: iㅣ, eㅔ, äㅐ, üㅟ, öㅚ, and 5 back ones: uㅜ, oㅗ, ŭㅡ, ŏㅓ, aㅏ, where ŭ and ŏare unrounded versions of u and o. (There are also 9 diphthongs, which are a recent addition and are disregarded here.) Four of the five front vowels, e, ä, ü, and ö, are contractions of the Early Middle Korean diphthongs ey, ay, uy, and oy. They are innovations from Late Middle Korean and did not exist in Early Middle Korean. On the other hand Earley Middle Korean had a further back vowel, ɔㆍ (called arae a), which disappeared at the end of Late Middle Korean.
     This suggests that if nothing had happened between Early Middle Korean and Modern Korean, Early Middle Korean would have had a vowel set consisting of 1 front vowel (i) and six back vowels (u, o, ɔ, ŭ, ŏ, and a). This is a highly unlikely constellation.
     To remedy this the author offers three steps to bridge the gap between the vowel sets of Early and Late Korean. Following Lee Kim-Moon (1977), the author posits the vowel set

    				front			back
    high	*i ㅣ	*ü ㅜ (rounded)		*u ㅗ (rounded)
    middle		*ö ㅡ (rounded)		*o ㆍ (rounded)
    low			*ä ㅓ (unrounded)	*a ㅏ (unrounded)
for Early Middle Korean (called "Old Korean" here). This vowel set is confirmed in three ways:
1. it agrees well with loans between Korean on the one hand, and Old Chinese, Middle Mongolian, and Japanese on the other;
2. Middle Korean vowel harmony opposed the vowels u, o, and a to the vowels ü, ö, and ä, resp.;
3. it makes the shapes of the Hangul letters look well-paired. [DG: This is an anachronism; Hangul was developed for Late Middle Korean.]
     Next, a circular vowel shift is then postulated:
    EMK ü → LMK u
    EMK u → LMK o
    EMK o → LMK a
    EMK a → LMK ä
    EMK ä → LMK ö
    EMK ö → LMK ü
apparently [DG: this is the only way I can interpret the diagrams in the paper] implemented by the author by rotating the Hangul letters one step clockwise in the above diagram:
				front			back
    high	*i ㅣ	*ü ㅡ (rounded)		*u ㅜ (rounded)
    middle		*ö ㅓ (unrounded)	*o ㅗ (rounded)
    low			*ä ㅏ (unrounded)	*a ㆍ (rounded)
Now it becomes clear that the old front-back-based contrast of the two columns is being replaced by a roundness-based contrast, and indeed, as it says in the paper, "the old palato-velar correlation, opposing */ü ö ä/ to */u o a/, was restructured into a roundedness correlation, opposing /ǔ ǒ a/ to /u o ɔ/". This results in:
    			unrounded	rounded
    high	i ㅣ	ŭ ㅡ		u ㅜ
    middle		ŏ ㅓ		o ㅗ
    low			a ㅏ		ɔ ㆍ
And this is indeed the observed Late Middle Korean vowel set.
[DG: The author gives no arguments for the posited circular vowel shift, and the subsequent restructuring. I can't tell if this is science or just token manipulation, playing on the shapes of the Hangul and Latin letters (ö into ŏ, etc).]
     To widen the scope to the surrounding language families, and again following Lee, the author splits the *i into two i-s, obtaining a supposed pre-Old Korean vowel set:
    		front	back
    high	*i	*ı	unrounded
    high	*ü	*u	rounded
    middle	*ö	*o	rounded
    low		*ä	*a	unrounded
Similar analyses, including the vowel harmony, are then performed for Mongolic and Tungusic, resulting in vowel sets for proto-Mongolian and Common Tungus that are almost identical to that of pre-Old Korean. For Nivkh (called "Gilyak' here) and the Chukotko-Kamchatkan languages (called "Luoravetlan" here) no full analyses could be done due to lack of materials, but the vowel harmony systems are analysed in the same way. This suggests areal correspondences.

* Ligeti, L., A propos du rhotacisme et du lambdacisme, (in French: About Rhotacism and Lambdacism), Central Asiatic Journal, vol. 24, #3/4, 1980, pp. 222-250.
Very clear and detailed exposition of the situation.
The first to notice that Turkish z went to r in Old Chuvash was J. Budenz in 1871, who called the phenomenon rhotacism, after the same effect in Latin grammar, e.g. genus, generis. Ramstedt reversed the direction to rz, but not all r went to z, so he split the r in and (same for l), at the same time declaring and as palatalized. Still these are just hypotheses.
     The palatalization idea was based, among others, on the Hungarian word borjú `calf' < proto-Turkic (proto-Altaic?) *bur²āγū (hence proto-Common Turkic *buzaγu) where -γu is a suffix. However, when this word entered Old Hungarian it lost its γ: bur²āū and then it developed as follows: bur²āūburŏuburĕuburĭu, to finally yield Modern Hungarian borjù. This makes clear that the j in Hungarian (which is just the first letter of the second syllable) does not come from proto-Turkic. This is confirmed by the Hungarian word gyapjú `wool': there is no r or other palatalizing entity in the Turkish form yapaγu nor in the Old Bulghar form *ǰapaγu, and there is still a j in Hungarian.
     Given that the Hungarian l by itself already has the tendency to palatalize (the author gives 6 examples) unambiguous evidence of = λ from Hungarian is hard to achieve. Some evidence can be obtained from Hungarian words with /-lč-/, spelled lcs in Hungarian. We have Hungarian bölcsö ~ Uyghur biśik `cradle', but Chuvash has bešik. [DG: seems a good approximation of a voiceless l, if you're not used to hearing it.]

* Mabuchi Kazuo, Ri Toraei, Ōhashi Yōko, 『三国史記』記載の「高句麗」地名より見た古代高句歴語の考察, (Sangokushiki kisai no Kōkuri chimei yori mita kodai Kōkurigo no kōsatsu), (in Japanese: Considerations of ancient Goguryeo words from the viewpoint of the "Goguryeo" place names from `History of the Three Kingdoms'), in Studies in language and literature. Language 4, 1979, pp. 1-47.
Analyses the place names from the `Samguk Sagi'. Is the basis of Beckwith's work (2004).

* Tekin, T., Once More Zetacism and Sigmatism, 1979, pp. 20.
==== Summary of the rhotacism/zetacism debate ====
Some (but not all) words with r in Chuvash/Bulghar have z in Common Turkic, and some (but not all) words with l in Chuvash/Bulghar have š in Common Turkic. Two hypotheses suggest themselves:
     1. The rhotacism hypothesis: proto-Turkic had z and š in addition to r and l, and Chuvash/Bulghar underwent
• rhotacism, i.e. zr (a common sound rule, f.e. Latin flos `flower', Old Latin flosis → Classical Latin floris, and
• lambdacism, i.e. šl (rare, but see, f.e. Benedict, `A Note on Proto-Burmese-Lolo Prefixation' (1975));
     2. The zetacism hypothesis: proto-Turkic had two r-like sounds, r and , and two l-like sounds, l and , and Common Turkic underwent
• zetacism, i.e. z (a common sound rule, f.e. Latin cathedra → French chaise), and
• sigmatism, i.e. š (unusual, but there is proto-Semitic ɬ → Ge'ez ś and proto-Semitic ɬ → Hebrew s (the Hebrew `sin', `shin' with left dot); see Y. Blau, `Phonology and Morphology of Biblical Hebrew: An Introduction' (2011), pg. xii, 69).
Possible phonetic values for and have been suggested: a palatalized r (ŕ), and a palatalized l (λ) (by Poppe and by Ramstedt); a voiceless r () and a voiceless l () (by Serebrennikov); and rty and lty (by Pritsak).
     There is no doubt that the sound correspondence `Common Turkic z = Chuvash r' exists, but there is great disagreement over the direction of the sound law. Finding out which direction is the correct one matters for the Altaic hypothesis because there are several words with r in Chuvash and z in Turkic that connect Chuvash to Mongolian, where they have r. Under the zetacism hypothesis the r was proto-Turkic, the words were proto-Turkic, and because they also occur in Mongolian, it would be reasonable the assume that they were also in proto-Turkic-Mongolian and so they would support the Altaic hypothesis. Under the rhotacism hypothesis, however, the Mongolian words cannot be related to proto-Turkic, or they would also have had z, so they can only be loans from Chuvash, and are meaningless to the Altaic hypothesis.
     In short: what came first in proto-Turkic, the r (and it became a z in Common Turkic through zetacism, supporting Altaic); or the z (and it became r in Bulghar/Chuvash through rhotacism, being neutral to Altaic)?
==== End of summary ====
     The author quotes Róna-Tas (1970) as saying that Ancient Turkish was not a homogeneous language, and consisted of many dialects, some having r and others having z. The author comments that this view brings us nowhere: "the members of a given language family descend from or go back to a single proto-language in the past". And that language had either r or z. [DG: It seems likely that the existence of a homogeneous proto-language is a dogma.]
     After a historical overview of the publications on the subject (summarized above) the author supplies many examples of rhotacism, all discussed in detail. There are fewer examples of sigmatism, and it is found that Common Turkic ś sometimes goes back to , , , or , in addition to .
     Doerfer's suggestion that was an allophone of r before i is refuted. The sound transformation ŕz is hypothesized to have gone through an intermediate ź. The author lists 27 words that end in z and have a derivative with r in non-final position, e.g. sämiz `fat' ~ sämri- `to become fat'; the r is considered original, pre-zetacism. Similarly, four words with šl are given, e.g. äš- `to dig with hands' ~ äliγ `hand'. But this cannot be the whole story, as there are also word pairs in which one word ends in z and the other ends in r. Three such pairs are given, e.g. qoŋuz `beetle' ~ qoŋur `chestnut'. This proves that r and are different phonemes, not allophones.

* Ramsey, S.R., Accent and Morphology in Korean Dialects: A Descriptive and Historical Study, Society of Korean Linguistics, 1978, pp. 264.
Not available. No abstract available.

* Lee Ki-Moon, Geschichte der koreanischen Sprache, (in German: History of the Korean Language), 1977, pp. 326.
German translation of 개정 그거사 개설 (`Revised Introduction to the History of the National Language'); the original was written between 1961 and 1972. The translator (Bruno Lewin) confessed in the introduction that the translation was difficult and had required repeated questions back to author and the Japanese translator.
     The book was originally written for Korean students of linguistics, who were fluent in Korean and Chinese. Both the author an the translator have spent considerable effort to provide additional information to help the Western user, but quite often Chinese text fragments have only been transliterated, not translated.
     The book consists of nine chapters. After the introduction, three chapters are concerned with the relation of Korean to the other Altaic languages; the languages in the first millennium AZ from which Korean arose; and the writing systems used for Korean. The further five chapters cover Old Korean (= the language of Silla); Early Middle Korean (still pre-Hangul); Late Middle Korean (after 1443); New Korean (until 1900); and Modern Korean.
     The book is very information-dense. Only a few of its features have been summarized here. It is telling that a significant number of sections end in "This requires more research."
     After pointing out that typological similarities cannot be used as evidence of a genetic relationship, the author lists 6 such similarities between the Altaic languages:
1. vowel harmony;
2. restricted consonant set in word-initial position;
3. agglutination;
4. absence of vowel or consonant alternations;
5. no relative pronoun, no conjunctions;
6. presence of a converb construction.
None of the surrounding languages (Indo-European, Uralic, Chinese, Paleo-Siberian) have all these properties, although Uralic comes close, lacking only converbs. The final conclusion of this chapter is that Turkic, Mongolic, and Tungusic definitely come from proto-Altaic, but if Korean is related at all, the Puyŏ language from which Korean derives, can at best have been a sister language of proto-Altaic.
     The author envisions the following vowel shifts:

	Old	Early	Late	New	Modern
	Korean	Middle	Middle	Korean	Korean
ㅏ	a	a	a	a	a
ㅓ	ä	e	ə	ə	ə
ㅗ	u	u	o	o	o
ㅜ	ü	ü	u	u	u
ㅡ	œ	ə	ï	ï	ï
ㅣ	i	i	i	i	i
ㆍ	ɔ	ɔ	ʌ	ʌ	--

     In the development of Old Korean (= the language of Silla) from Altaic the Altaic voiced and voiceless stops coincided. The aspirates started from hs elsewhere in words. This process started with kh, as with kho `nose' from koh and khar `sword' from karh. The basic grammatical endings -r, -n, and -m were already present in Old Korean.
     At the end of a word stops are pronounced without release (called "implosives" here) and -r changes to -l; both are the same effect.
     The text of the poem Ch'ŏyongga in the Silla original, the transcription in the Akhak-Kwebŏm (a book on musical instruments) in early Hangul, a version in modern Hangul, and the translation is presented on page 92. The author spends more than two pages in detailed criticizing the transcription of the four-line poem.
     The great insight of Hangul was that the the set of initial and final sounds of a syllable was (roughly) the same and could be represented by the same set of signs, unlike in Chinese.
     The Korean Vowel Shift is a persistent research line in Korean linguistics, and because Lee (1977) is the first major publication that spends considerable attention to it, this seems a good place to collect some notes on it.
     Talking about vowel shifts requires care: one should concentrate exclusively on the vowel and not be led astray by spelling. Ideally one should write all words concerned phonetically, but that is bothersome to do and many authors use traditional spelling. So when Lee writes that the was pronounced œ in the 13th century (pg. XII), when Hangul had not even been invented yet, I assume that he means that words that were written with in the 15th century, when there was Hangul, were pronounced with œ in the 13th century. This only works when all words with in them shifted their in the same way. We know that this is not true: the ㆍ disappeared by changing into at least three different vowels. So we are on thin ice here.
     Another confusing factor is the Yale romanization of Korean, which makes a number of weird assignments:
u is used for (normally ı), while wu is used for (normally u);
o is used for (normally ɔ or ʌ), while wo is used for (normally o); and
• most confusingly, the e is used for the ㅓ, which has been ascribed many values: ä, e, ə, ʌ, etc.
So it is of paramount importance to know if you are reading Yale romanization or phonetics.
     Much of what we know about the sounds of Old, Early Middle, and Late Middle Korean derives from Middle Mongolian, which is known from the 13th century onwards. Middle Mongolian was written in, Uyghur runes, Chinese characters, and Arabic script, with some texts available in more than one script.
• Based on all four scripts, Rybatzki (in Janhunen, `The Mongolic Languages' (2003), pg 63.) finds that Middle Mongolian had three back vowels, a, o, and u, three corresponding front vowels, e, ö, and ü, and a neutral vowel i, all pronounced as in Turkish or German.
• Based on the Arabic orthography, Saitô (see `The Vowel System of ...' (2007)) finds eight vowels: four high: i, ü, ï, u, three middle: e, ö, o, and one low: a, with length distinguished.
This may be indicative of how sure sound determinations of dead languages are. Still the phonetic values of the Middle Mongolian vowels are considered to be fairly certain.
     (The use of ö and ü as /œ/ and /y/ in Middle Mongolian should not be confused with their present-day pronunciation. Modern Mongolian has four back vowels, Cyrilic о, ө, у, and ү, usually (and quite reasonably) transcribed as o, ö, u, and ü. But these romanizations do not have the expected sound values; in particular ö and ü do not sound as in Turkish or German. It is easy to to assume that they are the palatalized counterparts of o and u, as they are in Middle Mongolian, but they are not. The modern pronunciation is:
     Cyrill.  Roman     IPA  tongue root
	о	o	[ɔ]	-ATR
	ө	ö	[o]	+ATR
	у	u	[ʊ]	-ATR
	ү	ü	[u]	+ATR
where the back vowels are distinguished by tongue root position. This shows us that Middle Mongolian also underwent a vowel shift. [DG: There is considerable variation in Mongolian pronunciation: for өрөө `room' the internet provides [ʊrɔ:], [oro:], [ərə:], etc. But the internet probably provides more social variety than was present in the writers of Middle Mongolian.])
     End of the notes about vowel shifts and back to Lee (1977).
     There are two roads to determining the vowel set of Old Korean: 1. from recordings in Chinese characters; 2. starting from the Early Middle Korean vowel set and work one's way backwards. With some hesitation the author arrives at the seven-vowel set for Old Korean (pg. 81)
    i    ü    u
      ö    ɔ
      ä    a

For Early Middle Korean the author arrives at a very similar constellation (pg.114):
    i    ü    u
    e    ə    ɔ
This involved the vowel shifts öə, äe and some repositioning.
     The author observes that the Middle Mongolian vowels from loans to Middle Korean are represented in Hangul as follows (pg. 113):
    a    ㅏ
    o,u  ㅗ
    e    ㅓ
    ö    ㅟ
    ü    ㅜ
    i    ㅣ
The ㅡ and the ㆍ (arae a) were not used because Middle Mongolian had no ɨ or ɔ.
     These loans entered Korean in the 13th century, so these correspondences represent the situation in the 13th century. As an additional argument the author puts forward that in the 15th century o and u were available (as ㅗ and ㅜ), so there would have been no need to map both o and u onto the same ㅗ. This means that, for example, in 13th century Mongolian there were words with ü that were copied into Old Korean with ü, where this ü changed into u in the course of the following two centuries, where it would be then recoded in Hangul as ㅜ.
     [DG: In the light of the Old Korean vowel set presented above, the mappings here are actually quite easy to explain.
1. Old Korean did not have an o, so it is reasonable that it is perceived as some kind of u.
2. Old Korean ä is the obvious candidate for the Middle Mongolian e.
3. The correspondence ö ~ ㅟ is surprising and the author only remarks that "apparently no better fit was available", which is strange because both Old Korean and Middle Mongolian had an ö. Now suppose these ös matched in the 13th century, and the Old Korean ö arrived in the 15th century as ü as part of the vowel shift. This ü could then only be written as a palatalized u, ㅟ.
     This produces a next to seamless explanation of the relation between Old and Middle Korean vowels, which is impressive. But we should not forget that the Old Korean vowel set was determined partly by reasoning backwards from the Early Middle Korean vowel set, so there may be some circularity of reasoning present here.]
     The sounds of the vowels of Late Middle Korean are described in the `Hunmin jeong-eum haerye' using two features, degree of tongue retraction and state of mouth, as follows (pg. 134):
		  tongue retraction
		 much	little	 none
	round	 ㅗ o	 ㅜ u
 mouth	neutral	 ㆍ ɯ	 ㅡ ɨ	 ㅣ i
	wide	 ㅏ a	 ㅓ ä
The author disregards these descriptions because it is unclear how "to translate the terminology in the `Hunmin jeong-eum haerye' into terminology from modern phonology" (pg. 161) [DG: The authors of the `Hunmin jeong-eum haerye' were quite clear in their description, and there seems little reason not to believe them. They were there and they certainly knew what they were doing. The recent recognition that tongue root position plays a large role in East-Asian phonetics lends additional credence to their description. Also, the yang vowels are characterized by much tongue retraction, and the yin ones by little tongue retraction, which seems fitting. Although IPA does not work with "degree of tongue retraction", I have tried to find IPA symbols that would best represent the resulting sounds. Modern theorists may frown at trivalent features, but perhaps not everything is binary.]
     Basing himself on a list of correspondences between signs and Hangul signs from the `Sasŏng t'onggo', the author arrives at this vowel set for Late Middle Korean:
    i ㅣ   ü ㅡ    u ㅜ
           ə ㅓ    o ㅗ
           a ㅏ	   ʌ ㆍ
When one compares this to the vowel set of Early Middle Korean it becomes evident that a vowel shift has taken place. The vowel shift is identified as (pg. 163):
    EMK e → LMK ə
    EMK ə → LMK ɨ

    EMK ü → LMK u
    EMK u → LMK o
    EMK o → LMK ʌ
Many details of the vowel shift are presented.
     For a dissenting account see Ko Seongyeon `Tongue Root Harmony and Vowel ...' (2012).


* Doerfer, G., Proto-Turkic: Reconstruction Problems, 1976, pp. 59.
The author surveys the work done on the reconstruction of proto-Turkic, including his own, which is represented impartially in the third person singular. The work is commented upon, in the first person singular.
     Most interesting for the Altaic unity issue is par. 4.9 on page 30: `Rhotacism and Lambdacism'. For a summary of the problem see Tekin (1979). The issue is first mentioned by Räsänen (1949). A good example of the r ~ z correspondence is Common Turkic ekiz ~ Chuvash yəkər ~ Mongolian ikire ~ Manchu ikiri `twin'. Examples of l ~ š are much rarer.
     For the Altaic unity issue it is important to know whether the z/š are original or the r/l. If the z is original, the r in Chuvash is from zr, and the Mongolian and Tungusic words are loans from Chuvash (or some more unlikely scenario, e.g. there was zr in Mongolian too (no proof of that), and then the word was borrowed by Manchu, or so.
     Bouda (1947) has shown that l ←→ š is found all over the world and Serebrennikov (1960) has shown the same for r ←→ z, so from a phonetic point of view both are equally likely, and proof must come from further analysis.
     Even from above example alone (and there are many more) it seems obvious that the r is original and the z is secondary, but a definitive proof is elusive. On what amounts to reasons of symmetry in the proto-Turkic consonant set the author concludes that it is more likely that the r/l were original.
[DG: This is essentially Hamp's argument (1974).]

* Benedict, P.K., A Note on Proto-Burmese-Lolo Prefixation, Linguistics of the Tibeto-Burman Area, vol. 2, #2, pp. 289-291. 1975,
Shows that the Burmese-Lolo language Kanburi Lawa has (initial) l where proto-Burmese-Lolo has s, f.e. proto-Burmese-Lolo sik → Kanburi Lawa lak `tree' (the ia is regular: proto-Burmese-Lolo liŋ → Kanburi Lawa laŋ `neck'). This is an example of sl.

* Tekin, T., Further Evidence for Zetacism and Sigmatism, 1975, pp. 10.
Precisely that: 19 new etymologies with zetacism, 3 with late zetacism, and 17 with sigmatism.
     The paper closes with a very clear summary of the rhotacism versus zetacism situation in the Turkic languages.

* Hamp, E.P., The Altaic non-Obstruents, 1974, pp. 3.
By analyzing the features of the proto-Altaic consonants and how they affect their environment, the author concludes that the proto-Turkic/Proto-Altaic consonants and are a palatalized r (ŕ), and a palatalized l (λ), resp., fitting right in with the palatalized n (ń), and the palatalized d (ǰ).

* Poppe, N., Zur Stellung des Tschuwaschischen, (in German: On the Position of the Chuvash Language), Central Asiatic Journal, vol. 18, #2, 1974, pp. 135-147.
To be read.
     Interesting detail: the author estimates the difference between Chuvash and Common Turkic to be about the same as that between Niederdeutsch and Hochdeutsch...

* Doerfer, G., Zur Sprache der Hunnen, (in German: On the Language of the Huns), 1973, pp. 50.
Long-winded and wandering paper in narrative style. Much space is allocated to refuting theories of other researchers, in particular Altheim (1948) and Pritsak (1954).
     There is a theory that the Huns were the Hiung-Nu from the Chinese records, but not much is known about their language: a number of words and a ten-syllable oracle saying ("couplet"), purported to mean "Lead the army out, catch the army leader", all recorded in uninformative Chinese script. In 1963 the author (in `Türkische und mongolische Elemente im Neupersischen, Teil I') demonstrated how poly-interpretable this couplet is by interpreting it as Akkadian...
     The Hiung-Nu did not speak a Turkic language: 1. most Hiung-Nu words do not have a Turkic interpretation; 2. any similarities can be explained as Hiung-Nu (3rd century) → Old Turkish (8th century) loans.
     Identifying Hiung-Nu as a Yeniseian language is unconvincing: 1. only 12 out of the 190 Hiung-Nu words identified by Pulleyblank (1962) show similarity to Yeniseian; 2. comparing 3rd century Hiung-Nu to 17th century Yeniseian is one equation with two variables; 3. any similarities are better explained as Hiung-Nu → Yeniseian loans.
     The Hunnic realm consisted of many peoples. Only 3 nouns of Hunnic are known: strava `funeral meal (of Attila)', of Slavic origin; medos `a drink', of general Indo-European origin; and kamon `a drink', of uncertain origin but already present before the time of the Huns. So some Huns spoke a Slavic language.
     "Attila" means `daddy' (little father) in Gothic, and that's what it means. It does not make him a Goth: King George of England had a Greek name, the Turkish ruler Suleyman had an Arabic name. No mystery here. Tens of other "Hunnic" names are discussed in detail; many are rejected for not being provably a name of a Hun, or for being evidently Germanic, i.e. local. Two "strange" names are mentioned: Kουρσιχ and Xελχαλ. It seems that Hunnic names with a Germanic explanation have declination forms, whereas the rest does not.
1. We don't know what language the Hiung-Nu spoke.
2. We don't know what language the Huns spoke.
3. It can be assumed that they did not speak the same language.
4. It is likely that both spoke a isolated language with no survivors.

* Poppe, N.N., A New Symposium on the Altaic Theory, Central Asiatic Journal, vol. 16, #1, 1972, pp. 37–58.
Report on the 1969 Leningrad symposium "Problems of the Affinities between the Altaic Languages", commented with many pros and contras.
     Good source of counterexamples. E.g. some people claim that there are very few words with purported etymologies in all Altaic languages, but the author provides one complete page of counterexamples, consisting of basic vocabulary, and comments that it is hard to believe that all resulted from criss-cross borrowing.
     The pair bi/min `I, me' is considered very strong evidence of Altaic unity by everybody.
     Some people claim that the Altaic languages have few basic words in common; but so do English and Greek. Many other examples comparing the properties of Indo-European and Altaic are given.
     The author warns against rejecting cognates when the words are not an obvious match, as Clauson does. E.g. Turkish taš `stone' and Mongol čilaγun `id.' do not look alike, but when it is pointed out that -γun is a usual suffix in Mongolian, and that Turkish š corresponds to Mongolian l, they already look more similar (cf. Korean tol `stone').
     Melnikov observes that the Altaic languages have a tendency (called a "determinant" by him) towards simplification, in the order Tungusic < Mongolian < Turkish. An example is Manchu narxun ~ Mongolian nirai ~ Turkish yaz (from ńär² `new, fresh'. An impressive table of 6 structural features of increasing simplicity is supplied.
     Cincius emphasizes that, in order to investigate a word, all its related words should also be taken in consideration, for the result to be convincing. Two extensive examples are given.
     The paper is written in German-school style, with some paragraphs spanning more than a page.

* Róna-Tas, A., Some Problems of Ancient Turkic, 1970, pp. 22.
The author argues that Ancient Turkic was not a homogeneous language and consisted of many dialects, some having r and others having z. In inhomogeneous languages words can have one form in one dialect and a slightly different one in another. As a result the overall language has doublets. Examples: French, with chair and chaise; German with Reiter and Ritter; Hungarian with hajlik `to bend' and kajla `bent, awry'.
     This raises the question what a proto-language is exactly. Opinions differ:
• Bloomfield (1935) writes (in "Language"): "[a proto-language is] a kind of phonemic diagram of the ancestral form".
• Reformatskij (1960) writes: "[ ] a historical reality".
• Pulgram (1961) writes: "[ ]a methodological device, a formula, but not a real language".
     The rest of the paper is a detailed analysis of the distribution of doublets with r and z over the Turkic languages.

* Tekin, T., Zetacism and Sigmatism in proto-Turkic, Acta Orientalia Academiae Scientiarum Hungaricae, vol. 22, #1, 1969, pp. 51–80.
Common Turkic z and š correspond to r and l, resp., in Chuvash. Examples:
• Com. Turk. toquz: Chuv. taxar `nine'
• Com. Turk. täš: Chuv. čul `stone'
Turkish being the bigger language, it was first (19-th century) assumed that the Turkish z was original and had changed in Chuvash to r through rhotacism. But in 1922 Ramstedt realized that this rhotacism extended into Mongolian and Tungusic, which was absurd. So the conclusion was that the z in Turkic came from an r in proto-Turkic through zetacism.
     A few other theories have been put forward, and are rejected/refuted by the author:
• Pritsak claims that = rti, = lti; but this is unlikely because these clusters are not very stable.
• Biyishev claims that -r and -z are just different endings in different languages; but the author shows that Biyishev's own examples do not support this.
• Ščerbak claims that r and z are allophones of proto-Turkic *s; but proto-Turkic already had a z opposing the s, of which the author gives examples.
     The author offers 38 etymologies in evidence that Ramstedt's theory is correct, that no rhotacism occurs, and that zetacism/sigmatism occurs only at the end of a word.
     After analyzing of how zetacism proceeded though the Turkish languages, the author offers examples of "Late Zetacism", "Late Sigmatism", and of these processes in suffixes.

* Doerfer, G., Zwei wichtige Probleme der Altaistik, (in German: Two Important Problems in Altaics), 1968, pp. 19.
This is a partial response to Clauson's 1966 review of the author's 3-volume `Türkische und mongolische Elemente im Neupersischen', which, next to praise for the size of that work, also features phrases like "Dr. Doerfer is unhappily less at home in Turkish than Persian and Mongolian..." So the "Important" in the title may be sarcastic, and indeed the author treats the two problems somewhat dismissively (but very thoroughly). But there is a lot more going on here; the author is clearly annoyed by Clauson's aggressive anti-Altaic stance. Sociology of linguists must be a fascinating subject...
     The author's position to the question of Altaic unity is that words that look like very good candidates for proving this unity, either indeed prove it or are very very old loans, and we cannot tell which it is.
     The first problem concerns similar words in all three branches of Altaic, which in Tungusic start with f/p, in Mongolian with h and start without a consonant in Turkic. Such a word can be explained most simply by assuming that it was one word initially and started with p, that it remained p in Tungusic, became h in Mongolian, and disappeared in Turkic, through ph → nothing, a well-known sound rule. The author tells us that Clauson claims that these words were originally Turkic, started with a vowel, and were borrowed by Mongolian, where they obtained a secundary h-. The author gives short thrift to this idea.
     The second problem concerns eight Ottomanic words in which according to Clauson the proto-Turkic initial p had been preserved. These words start with b in all other Turkic languages and in older texts, and because all of them end in a voiceless consonant the author suggests that the words originally had b and that retrograde assimilation to the voiceless final consonant had devoiced the initial b to p. Also, in older Ottoman texts (in those that distinguish voiced from voiceless) some of these words occur with b.

* Martin, S.E., Lexical Evidence Relating Korean to Japanese, Language, vol. 42, #2, 1966, pp. 185-251.
Explicitly sets out to help prove the relationship of Korean to Japanese. The relationship of Turkic, Mongolic and Tungusic as Altaic is widely accepted. The proto-language reconstructed from Old Japanese, mainland Japanese, and Ryukyuan is assumed to be the language of the Yayoi culture, which brought rice cultivation to Japan.
     Many languages of North Asia have the following properties:
1. The main word classes are nouns and verbs. This is supplemeted with a limited set of particels and endings.
2. Adjectives are included in the verbs.
3. Nouns can occur without ending, verbs forms require endings. [DG: This goes away when we accept ∅ (zero-length) endings, e.g. for nominatives and imperatives.]
     Although the North-Asiatic languages share these traits, they are also occur elsewhere, e.g. in Burmese, and are not indicative of a relationship: there are very many languages and only a few language types, so duplicates must occur. Similarity of Korean words with words in many other languages (e.g. Korean and English: manh- `many' and myech `(how) much') can be explained by the fact that both Korean and English are rich in synonyms, making coincidences like these very likely. Our body of evidence must be so large that it excludes the possibility of such coincidences. Once that is achieved only two causes of similarity remain: borrowing and common heritage.
     A strong indicator of genetic relationship is "shared shape alternations": English drink - drank - drunk and German trinken - trank - getrunken proves directly that English and German are related. In Altaic we find a strong shared shape alternation in the 1sg and 2sg personal pronouns bi/min `I/me' and si/sin `thou/thee' which occur in all three Altaic languages. These do not extent to Korean and Japanese, although proto-Japanese *ban(u) `I/me' might be related.
     Discussing borrowing, the author makes a dubious remark: "if any borrowing is forced to occur far enough back in time [through many regular phonetic changes], it becomes indistinguishable from common heritage and is to be treated as such."
     Sound rules that work well on basic vocabulary items are a plus. Sound rules can get support from related languages outside the ones we are comparing. E.g., the rules described in this paper derive Kor. *tol `stone' and Jap. *isi ` stone' from a proto-form *dyoš. This is supported by Ewenki dyol and Turkish taš. The system of sound changes should be internally consistent.
     Similarity of words in both languages even with the same meaning carries no weight unless there are so many instances that coincidence is excluded. The problem is deciding when that is the case.
     Next the author discusses the difficulty of getting the construction of sound rules bootstrapped: start with obvious cases in basic vocabulary. Extending the sound rule set and the vocabulary into an interlocking system resulted in 102 etymologies with a perfect meaning and perfect fit; 80 etymologies with perfect meaning and partial fit; and 83 etymologies with perfect fit and partially equivalent meaning.
     In parallel to the development of the sound rules a phonetic system of the underlying mother-language, proto-Korean-Japanese, was created. This system is maximal, in that it preserves all differences between Korean and Japanese. For example the ɔ in Korean that corresponded to an u in Japanese is reconstructed as ɔ, but the one that corresponded to o becomes ɔ'. This resulted in a very rich vowel system, richer than that of Korean and far richer than that of Japanese. The consonant system features a short voiced series b, d, j, a heavily aspirated series px .. kx, including a bx, and an l and an r. There is a large number of medial clusters, many of them involving the l.
     The next 26 pages specify the 32 sound rules, most of them with several variants. For each sound rule four items are specified: the rule number; the reconstructed sequence for proto-KJ; the sequences for Korean and Japanese; and a list of all reconstructed proto-KJ words that use this rule, with their English translations. The paper closes with 27 pages of 320 etymologies.

* Poppe, N., Introduction to Altaic Linguistics, Otto Harrassowitz, Wiesbaden, 1965, pp. 212.
The title should be taken literally: "Linguistics", not "Languages". The book is a survey of the knowledge about -- not of -- the Mongolian, Manchu-Tungus and Chuvash-Turkic languages, with a dash of Korean thrown in hesitantly. It is written for readers who do already have considerable knowledge of at least one of the languages involved; its addresses American linguistic students explicitly. No details about the languages themselves are given, except incidentally.
     The first chapter introduces the 37 languages + their dialects, detailing some of their history, location and speakers. There is much emphasis on the phonetic properties that distinguish one language from its neighbors on the language tree. Special attention is paid to the various alphabets in which the languages are written, including tables of these alphabets. Short bibliographies are supplied for each language; the author regrets the fact that few of them are in English. The relationships between the languages are presented in interesting, oriental-looking drawings.
     Chapter 2 covers the history of the study of the three groups. Chapter 3 explains the Altaic theory as impartially as possible, although the author reveals that he thinks that those against have more to explain than those in favor; he is even sympathetic to the inclusion of Korean (and perhaps even Japanese, pg. 147).
     Chapters 4, 5, and 6 try to unravel the influences of these languages on each other (4), on other languages (6), and of other languages on them (5). Much use is made of sound correspondences to distinguish loans from common origin. Chapter 7 treats the common structural features; Korean is an equal partner here. Chapter 8 gives extensive tables of sound correspondences between the Manchu-Tungus, Mongolian, Chuvash, and Turkic languages, with many examples of each. It looks like a convincing demonstration of the relationship between all four groups (and therewith of the validity of the Altaic theory), although the author does not press the point. Korean is absent here.

* Sinor, D., Observations on a New Comparative Altaic Phonology, Bulletin of the School of Oriental and African Studies, University of London, vol. 26, #1, 1963, pp. 133–144.
Critique of Poppe, `Vergeleichende Grammatik der altaischen Sprachen, Lautlehre' (1960).
1. It is only about Turkic and Mongolic, with little bits of Tungusic and Korean.
2. Poppe tries to establish a proto-Altaic, based on reconstructed Turkic and Mongolic forms, plus unwarranted proto-Tungusic and proto-Koreanic forms. Argued Tungusic forms would have been available from Cincius (1949), who is ignored, and the Korean forms seem to be created just to fit the bill.
3. The middle members of a chain of derivations are sometimes completely hypothetical.
4. Words that occur sporadically in some languages are considered a sufficient basis for reconstructing proto-forms. Nobody would consider a word occurring only in Spanish and in a French-Pyrenean dialect as Indo-European.
5. All evidence is examined in the firm conviction that proto-Altaic exists.
6. Words alone are not sufficient to establish a proto-language: on the basis of words alone one cannot decide if English is a Germanic or a Romance language. The real proof must come from morphemes. These are treated by Ramstedt, but his treatment is disappointing. Hopefully Poppe does better in his next volume, Vergeleichende Formlehre - Comparative Morphology.
     Next the author turns to his own ideas of Altaic unity.
1. Although Turkic and Mongolic have a large number of words in common, they do not share numerals [DG: they are supposed to have separated somewhere between 5000 and 3000 BZ; how interested in counting were people then?] and body parts [DG: that is worse; they did have body parts then].
2. There are many words for which Turkic t- ~ Mongolic t- ~ Tungusic t- holds; but there are also many words for which Turkic t- ~ Mongolic d- ~ Tungusic d- holds. Poppe believes that sound laws don't allow exceptions, and consequently he ignores the second group.
     There are smaller groups of words that exhibit similar but different three-fold rules. These irregularities can only be explained by borrowings from different dialects and/or at different times. The Altaic languages rather seem to be converging towards an Altaic unity than diverging from a common Altaic proto-language. Proto-Altaic is not only theoretical, it is also imaginary.

* Doerfer, G., Türkische und mongolische Elemente im Neupersischen, Teil I-IV, 1963, pp. 669.
(digitally available at
Page 96 of Vol. I discusses the (in)famous two-line Hiung-Nu/Chieh "couplet". It is written "phonetically" in Old Chinese. Several scholars have tried to find an interpretation of these 10 sounds recorded imperfectly in Old Chinese, in some language. The author supplies the following attemps:

Hiung-Nu	siôg   tyěg t`iei liəd käng, b`uokkuk g`yu t`uk täng
Ramstedt	sükä        ta-   li-  qiŋ,  bügüg         tu-  taŋ
Bazin		süg    täg- ti    id-  qan,  boquγi-  γ    tut- qan
Gabain		särig       ti-   lit- qan,  buγuγ    kö-  tür- kän
"Akkadian"	suddur Ta-        qid  ken,  puqqi    qud- di-  dam

The first line shows the Hiung-Nu original, the next 3 lines attempts by Ramstedt, Bazin, and Gabain. Their translations are given below. The author notices that the interpretations and translation vary so widely that anything could be interpreted into these 10 syllables, and to show how true that is he then supplies an interpretation + translation based on Akkadian.

Hiung-Nu: lead out army, catch army leader.
Ramstedt: go out to battle, catch the Bügü.
Bazin (as To-ba): send the army into battle, capture the commander.
Gabain: you will lead out the army, you will abduct the "Stag".
"Akkadian": the battle array will be set led out for sure, the overseer will be subdued.

No comment necessary.

* Pulleyblank, E.G., The Consonantal System of Old Chinese, (Appendix: The Hsiung-Nu Language), 1962, pp. 239-265.
Before the advent of Buddhism, the Chinese recorded very little foreign language material, but the language they recorded the most of was that of the Hsiung-Nu (= Xiong-Nu, 3nd century BZ - 3nd century AZ). And that amounts to a set of about 150 words (mostly hard to interpret names and titles) and a very short "couplet", all recorded in Chinese characters.
     Identifying the Hsiung-Nu words depends on the pronunciation of the Chinese characters at the time the text was written (about 600 AZ), and that pronunciation is exactly what was determined in the previous 100+ pages of this (two-part) paper. Using this information the author concludes that it is unlikely that the language is Altaic for the following reasons:
1. 9% of the words start with l; words in Altaic rarely start with l;
2. at least 14 of the words start with a consonant cluster, which is very rare in Altaic;
3. there are examples of *g and *d and hardly any of *kh or *th, but the author draws only vague conclusions from this.
The language is clearly not Sino-Tibetan, and there are no similarities with Indo-European (e.g. Tocharian). This leaves Yeniseian.
     Unfortunately very little Yeniseian material is available. All of it is recent (19th and 20th century), so there is a two thousand year time gap with Hsiung-Nu. The author presents perhaps a dozen Hsiung-Nu words with proposed relatives in Yeniseian, but each requires at least one page of text to make the relationship plausible. There is also a "couplet" in the Chieh language, which is supposed to be one of the languages of the Hsiung-Nu, with translation into Chinese. It consists of four words of 2, 3, 2, and 3 characters, resp. No attempt to match it to Yeniseian is made.

* Benedict, P., Thai, Kadai, and Indonesian: A New Alignment in Southeastern Asia, 1942, pp. 26.
Argues that a small subset of the Tai languages (Li on Hainan, Kelao in south-central China, and the Laqua and Lati languages on the China-Tonkin border) exhibits affinity with Indonesian, and calls this group `Kadai'. The hypothesis is based on the numerals and "a scattering of nouns". The similarity of the numerals, shown in a table, is not obvious, specially not when one replaces the very similar double columns of North and South Li and Kelao by single columns each.
     More recent research (Ostapirat, 2000) has shown that Benedict's languages do not form a separate family within Tai. [DG: Any vestiges of Indonesian in the Benedict languages may derive from both deriving from Austro-Tai, if it exists.]

* Hulbert, H.B., A Comparative Grammar of the Korean Language and the Dravidian Dialects of India, 1905, pp. 152.
The author, a non-denominational Christian missionary with no formal training in linguistics, learned Korean during his post in Korea from 1886 to 1905, without the use of grammar books. When he saw Caldwell's `Comparative Grammar of the Dravidian Languages' (1856), he was struck by the ease with which the text could be applied to Korean. The bulk of this book consists of feature descriptions quoted from Caldwell's book, each followed by an explanation of how perfectly the description matches the corresponding Korean feature. An example is the comparison construction:

    Dravidian: adei-parkkilum idu    nalladu
    Korean:    keu-boda       igeos  johda
`this-having-seen, that is-better' = `That is better than this', where indeed the word-by-word match is perfect. [DG: Dravidian from the original, Korean spelling modernized; the original was keu-poda i-gŭt ch'ota. Note ch'ota for chot'a.]
     The similarity of morphemes is also pointed out: Korean uses -t- for the past tense, as in gatta `went' (the t is actually the ss from the suffix -[a|eo]ss-) whereas Dravidian uses -d-; the oldest forms of Dravidian use g or k for the future whereas Korean uses get (actually -gess-); and Dravidian expresses negation with the particle a/al, whereas Korean uses an, and in both languages l and n are often interchangeable. But the author correctly identifies the m in -myeon `if' as coming from the substantivizer -m.
     Caldwell often describes a given feature for several Dravidian languages. On the basis of these the author concludes that Korean is most similar to Gondi, a Telegu language.
     After 13 chapters of structural similarities the author turns to the common lexicon, the highlight of which is formed by the 1st and 2nd person pronouns: Drav. nan ~ Kor. na `I', and Drav. ni ~ Kor. neo/ne. Another 83 Dravidian-Korean cognates are suggested. Examples are (with author's argumentation):
1. `Moon' Drav. tingal. Korean has shortened this to tal.
2. `Rain' Drav. pey. The Korean pi is nearly identical.
3. `Tree' Drav. manu. K. namu. Here we find a striking case of metathesis.
There is also a ingenious attempt to demonstrate similarities between the Dravidian and the Korean numbers.
     This book is interesting not because it emphasizes the structural similarity between Korean and the Dravidian languages, but because it allows us an unfiltered look at the Korean of around 1900. Several things can be noticed:
1. Many of the finite verb forms are in the blunt speech style: hae-bo-gess-so `will try', dar-a-na-o `runs (away)'.
2. The copula is -ilda rather than the present-day -ida [DG: but the latter still has an emphatic form -iloda].
3. Several words have a where today they have eu, often after o or a: onal vs. oneul `today', pat-nan-da vs. pat-neun-da `he receives' (but also hageonal vs. hageoneul).
4. The syllable dyeo has not (yet) contracted to jeo.
5. The form yeogui is used instead of yeogi `here', etc.
     Two interesting observations by the author:
1. A nasal is just a vowel pronounced through the nose. So if a language avoids surds (voiceless consonants) between vowels, it will also turn surds after nasals into sonants (voiced consonants).
2. Based on the formal correspondence between jug-eo ganda and `going to die' where `to die' is an infinitive, the -a/-eo form jug-eo should be called `infinitive' [DG: this may be the origin of that tradition].
     The book closes with 5 appendices in which the author builds a scenario explaining how the Koreans came from India to Korea. round 2500 BZ India was populated by the Turanian people, a people with Mongolian traits. When the Indo-Europeans entered India they caused large-scale migration. Those of the Turanians who stayed behind became the Dravidian peoples, and one of the other group moved to Ceylon, Indonesia, Formosa, and a chain of islands, to Korea. On the basis of this, the author produces 133 Austronesian words (from Efete) as possible cognates for Korean words. Needless to say, they are even less convincing than the Dravidian-Korean cognates presented in the body of the book.

* Budenz, J., Jelentés Vámbéry Á. magyar-török szóegyezéseiről, (in Hungarian: Report on the Turkish-Hungarian Word Lists of Á. Vámbéry), Nyelvtudományi Közlemények, #X, 1871, pp. 67-135.
The author noticed zr between Turkish and Chuvash, and called it "rhotacism" in analogy to Latin grammar; literature reference for Ligeti (1980).