Aisin Gioro (2012: 13) listed three readings for the Khitan large script character
1. ɔʊ (from 1996-2002 works by unspecified researcher[s]; she regarded this reading as erroneous)
2. ɑʊ (also from 1996-2002 works by unspecified researcher[s])
3. au (from Aisin Gioro 2004a)
These appear to be guesses influenced by Manchu abka [ɑpqʰɑ] 'heaven' and Jurchen
transcribed in Chinese as 阿卜哈 *apuxa (for [ɑpqʰɑ]?) and 阿瓜 *akwa (for a dialect variant**
[ɑʁwɑ] < *ɑʁβɑ < *ɑɢbɑ < *ɑqpɑ < *ɑpqɑ
with metathesis, voicing, and lenition: *ɑpqɑ?). These words and Nanai apqa (a loan from Manchu; Cincius 1975 I: 8) are isolated in Tungusic and could be loans from some Khitan-type language. So it is appealing to claim that a pre-Khitan *abuqa was borrowed into Jurchen before it was reduced to Khitan au. But that reading is purely speculative, though I am certain that the Khitan diphthong au is partly from *abu, since Khitan
<tau>/029 <tau> 'five'
corresponds to Mongolian tabun 'id.' (The Khitan diphthong au is also from earlier *au, as Khitan
<tau.lia> ~ <tau.lia> ~ <tau.lia> ~ <tau.lia>/029 206 189 <tau.lí.a> 'hare'
corresponds to Mongolian taulai 'id.' (The difference in final vowels eludes explanation.)
Kane (2009: 63) listed several more proposed readings of :
These are clearly guesses inspired by the Turko-Mongolic word for 'heaven': Old Turkic täŋri and Written Mongolian tngri. Neither Kane nor I accept them because of the lack of corroborating phonetic evidence: i.e., alternations between and other characters pronounced təŋ, etc.
Kane (2009: 63) mentioned and rejected a case of apparent alternation between and
If were <ɔʊ>, <ɑʊ>, or <au>, such an alternation would be plausible, but I know of no other evidence for the monophthongization of au or the diphthongization of o in Khitan.
If alternated with
161 <au> and/or 210 <áu>
I would be more inclined to reconstruct it as <au>. However,
without such alternations or any other phonetic evidence (e.g., Chinese
transcriptions) in support of any reading, I can only agree with Kane and transliterate semantically as <HEAVEN>.
*Three-digit numbers for Khitan small script characters are from Qidan xiaozi yanjiu (Research on the Khitan Small Script, 1985). There is no standard numbering for Khitan large script characters.
**The transcription 阿瓜 *akwa is from the Sino-Jurchen Vocabulary of the Bureau of Interpreters which contained no Jurchen script, so 阿瓜 *akwa may never have transcribed a Jurchen reading of
Perhaps literate Jurchen always pronounced that character as abka even though they may have used words like [ɑʁwɑ] in their native dialects. The Vocabulary of the Bureau of Interpreters may be a valuable source for one nonliterary Jurchen dialect. Yet another nonliterary Jurchen dialect underlies standard Manchu.
188.8.131.52:59: ASTERISKS AT AMARAVATI
My last post got me thinking about how I use (or don't use) asterisks to indicate reconstructions.
Here is the somewhat inconsistent logic behind my (non)use of asterisks on this blog so far. You may find exceptions in the thousands of entries I've written since 2002.1. Unattested and impossible forms from languages at any point in time
Marked with an asterisk: e.g., *ra is not a possible Tangut syllable since oral vowels must be retroflex after r-. (I specified "oral" because Tangut has the syllable riẽ with a nasal, nonretroflex vowel after r-.d
All of the tangraphs above are pronounced 2riẽ; there is no Tangut syllable 1riẽ.
Although Tangut has a huge vowel system, it lacks nasal retroflex vowels.)2. Unattested but hypothetically possible forms from languages at any point in time
Marked with an asterisk: e.g., Vovin (2010: 53) proposed that the Western Old Japanese genitive suffix -tu is from an Old Korean genitive suffix *-ci which is not attested in the very limited Old Korean corpus.3. Attested forms from modern languages
4. (Romanizations of) attested forms in (relatively) unambiguous scripts at any point in time
Examples: Russian, Manchu, Mongolian (even in the ambiguous traditional script - the vowels can usually be confirmed), Arabic (again, the vowels can be confirmed), Written Tibetan, Written Burmese.
Old Japanese and Middle Korean also fall into this category for me. There is no absolute consensus on how to pronounce man'yōgana and early hangul, but the reconstruction involved is not as heavy-duty as that of, say, Tangut or Middle Chinese. (See below. That didn't stop me from writing a full-length book on the reasoning behind my reconstruction of Old Japanese phonology, though!)
5. Romanizations of attested forms in extinct languages in ambiguous scripts
The Tangut, Jurchen, and Khitan scripts are "ambiguous" in the sense that there is no absolute consensus about how to read them. (There is a near-consensus about the Ming Dynasty readings of Jurchen large script characters, but there is still some disagreement on details, as one can see by comparing reconstructions in Jin Qizong's 1984 dictionary.)
I've long felt that an asterisk was redundant for Tangut, Jurchen, and Khitan, since all Roman letter readings in modern scholarship are always modern reconstructions. But by that logic, an asterisk would also be redundant for, say, Proto-Indo-European forms since PIE forms are also always reconstructions by definition. (A language with 'proto-' in its name is always a reconstruction. But not all ancestral languages are reconstructions: e.g., Latin is attested.)
I guess my inconsistency is based on the usage of previous scholars: e.g., I can't remember seeing Tangut written with asterisks, but PIE is normally written with asterisks. Kane (2009) generally wrote Khitan with asterisks* but wrote Jurchen both with and without asterisks (e.g.,
Kiyose's *degun and Kane's *deu 'younger brother'** on p. 129 and
on p. 54).
6. Romanizations of attested forms in early relatives of living languages in ambiguous scripts
For some reason I feel compelled to treat premodern Chinese and early Vietnamese differently from Tangut, Jurchen, and Khitan by writing it with asterisks: e.g.,
天 Middle Chinese *tʰen
𡗶 Pre-16th century Vietnamese *blời***
Tangut 1mə 'heaven'
Jurchen abka 'heaven'
Khitan au 'heaven' (Aisin Gioro 2004 as quoted in her 2012 paper; added reading 3.8.3:25)
I think it's because unlike TJK, Chinese and Vietnamese are still alive, and I want to graphically distinguish premodern and modern forms.
3.8.1:51: I would also star reconstructed premodern readings of extended sinographies other than Vietnamese nôm: e.g., Zhuang sawndip.
7. Transliterations (as opposed to phonetic reconstructions)
No asterisk, even if the transliterations are based on modern guesses. Strict transliterations are in angled brackets.**** Periods indicate divisions between characters of a block or word: e.g., Khitan 'dog' below.
3.8.2:08: ADDENDUM: Italics at Amaravati
I use italics to indicate
- emphasis of whole words (but I prefer boldface to draw attention to parts of a linguistic form)
- book and movie (but not article and episode) titles
- that something is a linguistic form: e.g., the word word and its PIE ancestor *werdho-
This usage is not universal: e.g., Vovin (2010) did not italicize reconstructions.
Phonetic transcriptions are never in italics; brackets are sufficient to indicate they are linguistic forms.
- that a word is not English: e.g., man'yōgana (but not hangul which has an entry in Merriam-Webster and is therefore not italicized when not discussed as a word: e.g., in section 4 above.)
*3.8.0:36: Kane (2009: 2) wrote the Khitan words from the History of Liao in a mix of reconstructions (e.g., ordo) and modern Mandarin readings of Chinese transcription characters (e.g., 捏褐 niehe instead of *ńiqo for 'dog'); none are preceded by asterisks. He listed various interpretations of the Khitan small script spelling
for 'dog' without asterisks: e.g., nəxi.
**3.8.2:16: The Jurchen large script characters are <deu.un> and correspond to Kiyose's *degun. (Oddly Kiyose reconstructed the individual characters as *deu and *un without *-g-.) Kane's *deu might have been written with the single character
by Jurchen literate in their own script.
The Jurchen small script character for 'younger brother' is clearly related to Chinese 弟 'id.' and the Khitan large script character
The Khitan large script character 弟 that is an exact lookalike of Chinese 弟 may also mean 'younger brother'.
***2.28.2:08: The Roman letter spelling blời [ɓləːj] is first attested in the 17th century. The earliest non-Vietnamese phonetic evidence for the reading of 𡗶 is the c. 16th century Chinese transcription 雷 *luj. A *b- did not 'grow' on the word by the 17th century; the absence of *b- in the Chinese transcription reflects the absence of *bl- in Chinese. Comparative evidence points to Proto-Vietic *bl- with a true *b- (i.e., not an implosive *[ɓ]; in Vietnamese, Proto-Vietic *b- devoiced to *p- which became *[ɓ]).
****2.28.2:54: I could regard my transcriptions of Written Tibetan and Written Burmese or even Middle Korean as transliterations, but I reserve angled brackets without italics for transliterations that greatly differ from a phonetic transcription: e.g., <ekāḥ> for Thai เกาะ [kɔ̀ʔ] 'island'. (See "An Aw-ternative Spelling in Lao" on the use of the graphemic combination <e> ... <ā> for o. That combination is not evidence for *e ... ā as a source of o-type vowels in languages written in Indic scripts. Thai [kɔ̀ʔ] was borrowed from an earlier Khmer កោះ <ekāḥ> *kɔh, partway between an even earlier *koh and modern [kɑh].)
184.108.40.206:59: AS IF I'M SINKING IN MO-DERN TRANSCRIPTIONS
I made a 'monstrous' mistake in "Dorga" because I mistook a modern reconstruction in Chinese characters for an ancient transcription of Khitan.
It is easy to confuse the two because they can be mixed interchangeably. For instance, Li Fanwen's 2008 Tangut dictionary has a Chinese character transcription for every single Tangut character regardless of whether those characters have known Tangut period Chinese character transcriptions.* The transcription 没 mo 'to sink' for
3513 1mə 'heaven'
is taken from the Timely Pearl (1190), but the transcription 如 ru '(as) if, be like' for the following tangraph
3514 2rioʳ (second half of 2257 3514 1tiọ 2rioʳ 'to infect')
is a modern creation. (Tangut r- was transcribed with *l-sinographs in the Timely Palm, since 12th century northwestern Chinese lacked an *r- that is now present in standard Mandarin. I do not know of any premodern transcription of 3514.)
Although there may be clues to determine which is which (e.g., the un-12th century use of Mandarin r-sinographs for Tangut r-syllables), it may be difficult to distinguish them at a glance.
My policy has been to ignore the sinographic transcriptions in modern sources like Li (2008) and Shi et al. (2000); although some of those transcriptions may match those in the Timely Pearl, I didn't want to take the risk of erring the way I did last night by building historical hypotheses atop modern inventions.
*3.7.1:03: I don't know what the purpose of all these transcriptions is. Is it to give the Chinese reader a ready-made phonetic equivalent for each tangraph that can be used in lay writing? That would make it the sinographic equivalent of my lay transcription of Tangut: e.g., 没 for 3513 is like my my and 如 for 3514 is like my ror. 没 and my, etc. are meant to give nonspecialists a vague idea of how the tangraphs were pronounced.
Jin Qizong's 1984 dictionary of Jurchen generally only lists Ming Dynasty Sino-Barbarian Glossary transcriptions. (There is at least one interesting exception that I want to discuss later.) I think a Tangut dictionary could likewise either only list Timely Pearl transcriptions. If it is absolutely necessary to have a sinographic transcription for all entries with known Tangut readings**, Timely Pearl-style transcriptions could be added with some sort of marking to distinguish them from genuine 12th century transcriptions.
The problem of artificial readings plagues even dictionaries of modern languages. Chinese, Japanese, and Korean sinographic dictionaries all have fanqie-based readings of obscure characters that were never pronounced in modern times. Those readings are not distinguished from readings that have been orally transmitted over centuries; the latter may not quite match the modern readings of fanqie spellers chosen long ago: e.g., 奪 has the Guangyun fanqie 徒活切 (Sino-Korean to + hwal), but its Sino-Korean reading is thal with unexpected aspiration, not *tal. (*tw- is not possible in modern Sino-Korean.)
**3.7.1:56: There is a small number of tangraphs whose readings are wholly or partly unknown: i.e., characters without entries in native Tangut phonological dictionaries (e.g., 1555, 6020, 6033) or only have a known initial category (e.g., 1860 which had a liquid class initial or 1911 which had an alveopalatal class initial. Obviously it is not possible to transcribe these readings in sinographs. Unlike Roman letters, sinographs cannot represent initials without finals outside a fanqie context, much less a generic initial class: e.g., there is no sinographic equivalent of capital L- for 1860 or CH- or 1911.
220.127.116.11:41: DORGA (CORRECTED)
(18.104.22.168:21: I thank Andrew West for pointing out a huge error in this entry. Instead of deleting it or revising it without comment, I will keep the original text up and add corrections in bold.)
I thank Andrew West for reminding me that the Khitan large script spelling
corresponds to the name transcribed in Chinese as 多羅里本 *tɔlɔlipun. (See the name in context here.)
No, 多羅里本 is a modern reconstruction, not a Liao Dynasty transcription. It would have been read as *tɔlɔlipun in Liao times if such a transcription had existed then, but its real purpose is to represent Cong, Liu, and Chi's (2003: 50) reconstruction *dorlipun in Chinese characters that are pronounced Duoluoliben [twɔlwɔlipən] in modern Mandarin. Although 多羅里本 is not marked with an asterisk, it is no more real than its Roman letter equivalent *dorlipun with an asterisk.
For years I had only seen pages 54 and 55 of Cong, Liu, and Chi (2003) on Andrew's site. I did not see the rest of the article until tonight; it clearly states,
'We might as well use the Chinese characters 多罗里本 to transcribe it [the Khitan large script name ].'
罗 is the simplified form of 羅 which appears in the glosses at the end of the article.
I am surprised the name was not written in Mandarin as 多兒里本 Duoerliben [twɑɻlipən] with a use of 兒 for *r that is clearly modern. (The Liao Chinese reading of 兒 was something like *ʐɨ without an *r.)多 羅 was preferred because it was more archaic-looking. Perhaps
Aisin Gioro (n.d.) preferred to reconstruct the name as 奪里不, pronounced Duolibu [twɔlipu] in modern Mandarin. I do not know how she would have written that reconstruction in Roman letters. My guess is something like *dorlibu. She reconstructed the reading *dor for the Khitan small script character 'seal' in this 2012 article.
He also pointed out an alternate, shorter transcription 奪里本 *tɔlipun of what may be the same name in the History of Liao.
The History of Liao was written two centuries after the fall of the Liao and its spelling 奪里本 may not have been accurately copied from earlier sources. Nonetheless it is not a reconstruction; it is the only (near-)contemporary transcription that I have on hand.
The interpretation of all three Khitan large script characters is problematic, but let's focus on the first in this post. Given the following -
- Chinese voiceless unaspirated obstruents (e.g., *t, *p) correspond to Khitan consonants that I transcribe as voiced obstruents (e.g., d, b; their exact phonetic values are uncertain)
This is why I am puzzled by Cong, Liu, and Chi's decision to reconstruct the name as *dorlipun with *p rather than as *dorlibun with *b.
- Chinese *ɔ corresponds to the Khitan vowel that I transcribe as o (whose exact phonetic value is uncertain)- Chinese *l corresponds to the Khitan consonants that I transcribe as r and l (since Liao Chinese had no *r)
- the Khitan large script character means 'seal' in other contexts
- the Jurchen large script character <doron> for 'seal' was similar: ~
- the Manchu word for 'seal' was ᡩᠣᡵᠣᠨ doron
- I reconstruct the following scenario:
- Khitan may have zero corresponding to -n in Jurchen and Manchu: e.g.,
<m.ri> mori : Jurchen and Manchu morin 'horse'
- The Khitan word for 'seal' was dor(o)
- The Khitan large script character for 'seal' was used as a phonogram <dor> in unrelated words: e.g., Dorlibun, glossed in the History of Liao as 討平 'pacify'. -bun is a past tense marker, so maybe the named means 'Pacified'. Khitan dorli- may be cognate to Mongolian daruɣu 'peaceful'.
- This dor was transcribed in Chinese as 多羅 *tɔlɔ or simply 奪 *tɔ because there was no Chinese syllable *tɔr or even *tɔl.
It was never transcribed in Chinese as 多羅 *tɔlɔ in Liao times. 奪 *tɔ is the only known premodern transcription.
It is tempting to think that 奪 was chosen to transcribe Khitan dor because it had a final *-r in Chinese. It certainly did have a final *-r when it was borrowed into Sino-Korean around the eighth century (its prescriptive Middle Sino-Korean reading is ttwarʔ and its modern Sino-Korean reading is thal*), but Khitan loanwords from Liao Chinese indicate that *-r was gone by the 11th century. It is possible but unlikely that the transcriber knew of an archaic *-r reading for 奪, even though he himself normally pronounced the graph as *tɔ.
- The Khitan word was borrowed into Jurchen with an -(o)n suffix added
- Manchu inherited the word from its source dialect of Jurchen (which was not the dialect underlying written Jurchen or the varieties recorded during the Ming)
Revising Doronga to fit that scenario still results in a name appropriate for a Japanese giant monster: Dorga.
*The width of the gap between prescriptive and actual 15th century Sino-Korean is unclear. The actual reading of 奪 may have been something like *th(w)arʔ.
tt- might have been [d] rather than the tense tt- of modern Korean.
The initial th- of the modern reading thal is irregular; prescriptive tt- normally corresponds to actual t-. There is no Korean-internal reason for the aspiration, so I wonder if it reflects a northeastern Late Middle Chinese dialect with *tʰ- (< Old Chinese *s-l- with a prefix) instead of standard *d- (< Old Chinese *l- sans prefix). Unfortunately, I don't know of any modern Mandarin dialect with an aspirated initial in 奪. Tangut period northwestern Chinese had *tʰ-, and Gan and Hakka to the south still have tʰ- today, but none of those languages are likely sources for Sino-Korean thal.
sounds like the name of a Japanese giant monster. Ha, in fact, thanks to Google, I learned there was a Doronga in episode 10 of Golden Bat (1967).
Doronga is a play on Japanese 漫画 manga.
was the Jurchen word for 'seal' (cf. Manchu ᡩᠣᡵᠣᠨ doron 'id.'). The word may have been borrowed from Khitan
whose large script character (left; the small script character is on the right) is surely somehow related to the Jurchen large script character, though whether the relationship was direct or indirect** is unknown.
The Khitan large script character for <doro> in turn coincidentally resembles 立 over 儿, a variant of 漫 màn 'unrestrained' (the first half of Japanese 漫画 manga) that I found in Thesaurus Linguae Sericae (TLS). The nineteen variants in TLS fall into six categories:
1. Slight graphic variation of the phonetic component: e.g., 澷 with 方 instead of 又 on the bottom of 曼 màn
2. Different phonetic component: e.g., 澫 with 萬 wàn < *m- instead of 曼 màn
3. Different semantic component: e.g., 墁 with 土 'earth' instead of 氵 'water'
4. Different phonetic and semantic components: e.g., 㙢 with 土 'earth' and 㒼 mán instead of 氵 'water' and 曼 màn
5. Completely different semantic compounds:
𣸞 < 平 'level + 水 'water' x 2𧽾 < 走 'run' + 散 'scatter'
6. Completely different and opaque: what appears to be (but isn't necessarily) 立 'stand' over a pair of 儿 legs
The sole example of this last category is the most intriguing. What is the reasoning behind that shape? Where is it attested?
*3.5.0:21: <doro> is Kane's (2009: 114) reconstruction of the Khitan word for 'seal'. I do not know of any alternate phonetic spellings or transcriptions that confirm his reconstruction. More on that word here.
**The Jurchen large script may be a direct offshoot of the Khitan large script, or both scripts may be offshoots of an earlier Parhae script (Janhunen 1994).
The Russian word for 'three' is три tri. So why isn't тройка trojka 'troika' *трика trika? I assume it is derived from the collective numeral трое troe.
In Proto-Indo-European, the vowel of 'three' varied between *e, *i, and zero: e.g.,
*trí-ns (masc. acc.) ~ *tréy-es (masc. nom.) ~ *tr̥y-óHom (neut. gen.)
The basic stem was *trey-; when it lost its vowel, *y became *i (i.e., syllabic y) before a consonant-inital suffix and *r became syllabic *r̥ before *y and a vowel-initial suffix. There was no *o-grade stem *troy-. Is the o of Russian troe, Polish and Czech troje, etc. inherited from Proto-Indo-European or a Slavic innovation?
I still don't understand why the feminine stem of 'three' was *tisr-. Was it *tri-sr- at an even earlier stage? Today I discovered that Bachmann (2005: 71) also derived *tisr̥- from *trisr̥-. The suffix *-sr- may also be in the feminine noun *swésōr 'sister'. But see Emmerick (1991: 293) for another explanation.
22.214.171.124:52: ON *PAːROL (OR *PAːROL ET LANGUES)
Having closely examined Whitman's Proto-Koreo-Japonic (PKJ) *zitör a couple of entries ago, I'd like to look at his PKJ *paːrol 'needle' in detail:
|Early Modern Korean||p||a||r||ʌ||r|
|Eastern Old Japanese||p||a||r||u|
|Western Old Japanese||p||a||r||i|
*p lenited to ɸ in Middle Japanese and is now h in modern Japanese and Shuri. *p lenited in languages to the east and west of Korean (i.e., Mongolian, Manchu, and Japanese) but not in Korean itself.2. *aː
Does Shuri preserve the long vowel that Whitman reconstructed in PKJ? Here are two different scenarios:
The 'yes' scenario: According to Whitman (1990: 528), PKJ *-r- was lost after a short vowel in pre-Old Japanese (Proto-Japonic?). If the long vowel of Shuri is projected back into Proto-Japonic, then *-r- could not have been lost after aː at the Proto-Japonic stage, and Japanese -r- is retained from PKJ *-r-. (See section 3 for the Shuri reflext of PKJ *-r-.)
The 'no' scenario: Whitman (1990: 531) proposed that the Shuri long vowel is the result of contraction (in Proto-Japonic?): i.e., Shuri aː is from PKJ *-aro- whose *-r- was lost after a short vowel. So where did Proto-Japonic *-r- come from? See section 6.
The low tone of Middle Korean à matches the *low tone reconstructed for *a in orthodox Japonic tone theory.
3. *r or *n
According to Vovin (2010: 96), nearly all Koreanic evidence points to *n, not *r. The exceptions are:
- Early Modern Korean parʌr, a hapax legomenon from the early 17th century (dating after Middle Korean pànʌ́r with -n-)
- Kyŏngsang dialect forms (paːl ~ pal) which might be from parʌr or panʌr
Whitman (1985: 149) regarded the first -r- of Early Modern Korean parʌr as "a later independent assimilatory development in Korean": i.e., n became r before another r. On the following page, he proposed the same assimilation in Japonic, presumably after PKJ final *-l became *-r:
*paːnor > *paːror (vowel length added*)
The title of this post is based on PKJ *paːrol from the appendix of his 1985 PhD dissertation. Is that *r supposed to be *n?
In the 'no' scenario (section 2), PKJ *-r- was lost in Japonic before short *a.
In Shuri, *-r- was lost before *i: *aːri > aːi. Shuri ari is from *are, not *ari:
ari 'that' < *are (cf. standard Japanese are)
but ai 'ant' < *ari (cf. standard Japanese ari)
Whitman (1985: 129) proposed the following correspondence:
PKJ *o : Middle Korean ʌ : Old Japanese u word-internally; a in long syllables
Mid *o is intermediate in height between high u at one end and lower mid ʌ and low a at the other. Its rounding is retained in Old Japanese u but lost elsewhere.
Eastern Old Japanese paru retains Proto-Japonic final *-u, but Western Old Japanese does not. See section 6.The high tone of Middle Korean ʌ́ matches the *high tone reconstructed for *u in orthodox Japonic tone theory.
I do not know why Whitman reconstructed PKJ *-l instead of *-r.
Only Korean retains a final liquid in the 'yes' scenario.
On the other hand, Proto-Japonic and mainland Japanese retained *-l as -r- in the 'no' scenario - but in that case, where did the following vowels come from? I deal with *-i in section 6, but there was no Japonic suffix *-u.
6. Japonic *-iEast Old Japanese preserved the bare Proto-Japonic stem *paːru (possibly sans vowel length), but other varieties of Japonic have a final -i that is a contraction of stem-final *u plus the noun suffix *-i.
What do I think?
Unlike Whitman ̣(1985, 1990) who believed that Korean and Japonic inherited 'needle' from a common ancestor or Vovin (2010: 96-97) who rejects Whitman's etymology and who may regard the words as lookalikes, I think Japonic *paːru may be a loan from Koreanic. Japonic speakers may have misheard Koreanic *n as *r or borrowed from a Koreanic variety that had assimilated *n to a final liquid or shifted medial *n to a liquid (cf. modern Korean [kollan] 'difficult' from 困難 *konnan). Japonic -ru may be an attempt to imitate a Koreanic final *-r (from a Proto-Koreanic *-rʌl) or *-rʌw (< *-rʌɰ < *rʌɫ < *-rʌl). This borrowing occurred before Japonic speakers left the peninsula for Japan since the word can be reconstructed in Proto-Japonic. (Later borrowings - most likely from Paekche - are confined to Western Old Japanese.) The Koreanic source language may not have been directly ancestral to Korean; its line could be extinct.
*Whitman mentioned vowel length in the text but did not indicate it in the protoforms on that page. The long vowel blocks *-r- loss in Japonic according to the 'yes' scenario (see section 2).