Austronesians in the Northern Waters?

The following lines are inspired by John Kupchik’s seminal article ‘Austronesian Lights the Way’ that appears in this volume of JEAL. It demonstrated beyond any reasonable doubt for the first time that there are reliable Austronesian loanwords in Japonic that reveal quite ancient and profound contacts, because without these profound contacts the borrowing of the names of the most basic celestial bodies, such as the sun and the moon, would not be possible. In my opinion, his article opened a new and an exciting direction in the Japonic historical linguistics. There are, however, two important differences between Kupchik’s article and the present one. First, while Kupchik mostly concentrated on the Amis language from Taiwan, and to a less extent on the languages of Philippines and other Western Malayo-Polynesian, my major focus is on the Philippines languages as potential donors, and much less on other Austronesian languages of the region. Second, while Kupchik looked mostly on mysterious words in the Omoro S ō shi , a collection of Old Okinawan and Amami sacred and folk poems (1531–1623 AD), this article focuses more on Old Japanese in particular and Japonic in general.


Introduction
Some general remarks are in order. Overall, I agree with Lyle Campbell, who always said that "languages should speak for themselves". This means that other humanities data, such as archeology, history, religious studies, etc., as well as genetics from the natural sciences, which is in great vogue nowadays, cannot tell us anything either in support or criticism of the purely linguistic data. Yet, sometimes these data may be interesting and worth looking at, as they can potentially point into a right direction. Therefore, before diving into historical linguistics and philology, let me offer some preliminary extralinguistic observations. First, the distance between Yonaguni, the last island in the Ryukyuan archipelago, and the coast of Taiwan is mere 108 km. The closest landfall in Taiwan also appears to be in the region where the Amis people live. Among the Taiwan aboriginal tribes, only Amis and Yami are true seafarers, with Amis being famous for their long boats well suited for distant ocean voyages. From this point of view, it is incomprehensible why Austronesians would sail only south, and not north, especially that in clear weather one can see Yonaguni from Taiwan and vice versa. We, of course, know nowadays that there was Shimotabaru culture in South Ryukyus between 2900-2000 BC, but the archeological evidence supporting its connection to a culture found in North-Eastern Taiwan at the same time is inconclusive, and in addition there is a hiatus in population found between 2000 BC and 900 BC (Pearson 2013: 71-81). In any case even 900 BC is too early for Japonic speakers to be present in Sakishima, as they would not be able to reach there before 1200 AD.

Austronesian Vocabulary in Old Japanese Makura-Kotoba
One of the natural sources for searching Austronesian vocabulary in insular Japonic are those permanent epithets in the Old Japanese poetry that are opaque. These permanent epithets are known in Japanese as 'pillow words' (makura-kotoba 枕詞). It must be mentioned that the majority of them are quite transparent, and can be clearly explained on the basis of Japonic itself. Then there are two other groups, a larger one based on Old Korean, and a smaller one that has Ainu origin. Finally, there is one more group that cannot be analyzed on the basis of any these three languages. I will try to show below that at least some of those, or parts thereof can have Austronesian etymologies. One of the most spectacular cases is the permanent epithet taratine n-ǝ that is used with the OJ word papa 'mother' , but never with its doublet omǝ ~ amo 275 Austronesians in the Northern Waters?
International Journal of Eurasian Linguistics 3 (2021) 272-300 Eastern branches, is limited to Formosan and languages of Borneo in WMP, but these languages do not have daraɁ 'blood' . This distribution, I believe, probably points to the Northern Philippines as a place of origin, which lacks honorific t-, but has all other features attested.
To the best of my knowledge, among all OJ texts, tara tine is found only in the Man'yōshū, the earliest and the largest Japanese poetic anthology. It is quite clear that in OJ we deal with a very old prehistoric loan, because the meaning of the permanent epithet tara tine was already lost in the eighth century AD. Some poems engage in a folk etymological analysis, spelling tara ti as 垂乳 'hanging breasts' (MYS 9. 1774, MYS 11.2368, MYS 11.2557, MYS 12.2991), but such an interpretation is obviously ungrammatical, because with the vowel verb tare-'to hang down' we would expect tar-uru ti 'hang. down-ATTR breast' or tare ti 'hang.down(NML) breast' , but under no circumstances *tara ti would be possible.
I believe that OJ tara tine means 'one's own birth mother' . It looks like it was an important distinction to make, because in all likelihood, other wives and possibly even concubines of one's father were also referred to and/or addressed as 'mother' . As far as I am aware, there is no textual evidence to support this claim, but there is strong linguistic evidence supporting it. Blood lines in Ancient Japan were extremely important, and the matrilineal descent was as important as the patrilineal, with the former replaced by the latter only after the Heian period (794-1192 AD). Originally, the maternal line seems to be more important than the paternal one. In the Nihonshoki and other WOJ and MJ texts we find the following compounds, consisting of irǝ + kinship term, which designate kinship relationship by blood: WOJ irǝ-pa 'one's birth mother' , attested in the Nihonshoki exclusively in kana glosses that cannot be earlier than the Heian period: イロハ (NS 1.114), (NS 1.12), (NS 1.13), etc. There is also one more attestation irǝ-pa (伊呂波) in the man'yōgana in WMS (2.14b), which is much more useful than the kana glosses, in spite of the fact that WMS is from the early Heian period, but in this case we apparently have a case of an orthographic conservatism. WOJ irǝ-se (伊呂勢) 'brother from the same mother' (Omodaka et al. 1967: 107), although, attested in the man'yōgana script (萬葉仮名) in the Kojiki (KJK 1.23a), and possibly in a gloss for the character 兄 in the Nihonshoki, although there is also an alternative gloss irǝ-ye for the same character. OJ se refers exclusively to an 'elder brother' , 'male beloved' , or 'husband' , and never to a 'younger brother' . Therefore, the motivation for Omodaka et al. to define this combination as 'brother from the same mother' rather than 'elder brother from the same mother' remains unclear. The examples they cite support only 'elder brother' . Cf. also WOJ irǝ-ndǝ ~ irǝ-tǝ 'younger brother or sister from the same mother' below.
WOJ irǝ-ndǝ ~ irǝ-tǝ (伊呂杼) 'younger brother or sister from the same mother' is attested in a completely phonographic spelling in KJK 2.13b and 2.15b. The meaning 'younger sister' might seem strange, but it is confirmed beyond any doubt by example (1) below: (1) KJK 2.13a-13b 有二女。兄名蠅伊呂泥 ……. 弟名蠅伊呂杼也。 There were two women. The name of the elder sister was Fly the Elder Sister …… The name of the younger sister was Fly the Younger Sister.
WOJ irǝ-ne (伊呂泥) 'elder brother or sister from the same mother' (Omodaka et al. 1967: 108). This time Omodaka et al. seem to be right, since this combination is applicable to both sexes. The completely phonographic spelling is attested in KJK 2.13a-13b, see the example above. WOJ irǝ-MO (伊呂妹) 'younger sister from the same mother' is attested only in partially phonographic spelling in KJK 1.42b, 2.28b, 2.31a, and 3.17b. In all these cases we have kana gloss モ mo to the right of the character 妹 'younger sister' .
The etymology of WOJ irǝ is obscure. To the best of my knowledge, it is not attested in any other variety of Japonic except WOJ. Such a limited distribution 277 Austronesians in the Northern Waters?
International Journal of Eurasian Linguistics 3 (2021) 272-300 is highly suggestive of a borrowing. But what language did this loan come from? Ainu, Koreanic and Austronesian have no word that could fill in this niche. Quite possibly, this might be a borrowing from some Jōmon period language. But certainly this hypothesis is impossible to prove or to disprove, so most likely it will remain just a speculation. There is a tiny chance that WOJ irǝ-may also be from Austronesian, cf. WMP Paku iraɁ 'blood' (Blust and Trussel 2020). But besides an obvious problem with a vocalism in the second syllable, Paku iraɁ (a reflex of PAN *daRaq 'blood') is isolated in WMP. And finally, seeing an Austronesian behind every bush is going to be more harmful than helpful. However, notwithstanding its source language, most likely irǝ means 'blood' , so irǝ-pa is probably the construction identical to tara tine.
To the best of my knowledge, tara tine 'mother by blood; one's own birth mother' is not attested in OOK. At the first glance, there might be, however, a fly in the ointment: in addition to twenty-four examples of tara tine 'mother by blood' there are also three examples of the permanent epithet tara tisi for the word papa 'mother' in the Man'yōshū: 多羅知斯 (MYS 5.886), 多良知子 (MYS 5.887),8 and 垂乳爲 (MYS 16.3791). But ultimately these three examples might also have a solution. Tara 'blood' has been already discussed above. But t-isi might be another Austronesian word: honorific t-, also discussed above + isi 'flesh' , 'meat' (Blust and Trussel 2020: *isi 5 Consequently, tara t-isi n-ǝ papa means 'a mother, who is [of the same] blood and flesh' . A textual example:

) Without seeing the eyes of [my] mother, (1) who is [of the same] blood [and] flesh.
Finally, there is one example of tara tune (足常) in MYS 11.2495. I suspect that it is simply a case of an irregular development in WOJ, probably caused by progressive assimilation from the front vowel i to the back vowel u under the influence of another preceding back vowel a. Another interesting permanent epithet is nama yǝmi-nǝ (奈麻余美乃) with unknown meaning, which occurs only once in MYS 3.319 as a makura-kotoba for Kapï (甲斐) province. The yǝmi-nǝ part is comparatively uncontroversial, as it is likely to be -nǝ, genitive case marker, and yǝmi in all likelihood reflects yǝmi in Tuku-yǝmi10 'moon deity' . The important thing is that it cannot be Yǝmï 'Hades' , which is very popular as a folk etymology, but the second syllable vowels in yǝmi and Yǝmï are very different.11 Sometimes Tuku-yǝmi 'moon deity' , 9 I selected only those words from Blust and Trussel's list have meanings 'human flesh' and/ or 'meat' . 10 There is apparently some confusion in WOJ texts between yomi and yǝmi in this word. Most importantly, there is apparently an example of raising o ⟩ u: Tukuyomi ⟩ Tukuyumi that never happens with the vowel ǝ (Omodaka et al. 1967: 466), so it is more likely that the vowel was o and not ǝ. But we should not also exclude a possibility of contamination. 11 Surprisingly, even some scholars support this folk etymology. E.g. Nishimiya dedicates a long discussion where he tries to circumvent this problem, although he understands

Vovin
International Journal of Eurasian Linguistics 3 (2021) 272-300 tuku-yo 'moon-lit night' mean 'moon light' . Therefore, it is not unconceivable that yǝmi could do the same. The first word nama, however, is unlikely to be Japonic: while there is WOJ nama 'fresh' , 'raw' , it does not fit here semantically.
It is always notoriously difficult to deal with hapax legomenoi like this one, so our first step will be to examine the context of MYS 3.319. It is a chōka, but just a few lines will suffice (8-10): (8) Preface to the poem 3.319

詠不盡山歌一首并短歌
A poem composed about Mt. Punzi with tanka envoys.

Commentary
WOJ Mt. Punzi is the same as MdJ Mt. Fuji, a dormant volcano and the highest mountain in Japan. the difference between kō-rui vowel /o/ and otsu-rui vowel /ǝ/ (1984: 169-171). Omodaka is much more straightforward, just noting that the meaning of nama yǝmi is unknown (1977.3: 239). 12 I did not include this or other poems in this mini-sequence here.
I glossed above nama as 'deep river' , because I think that the word reflects yet another Austronesian loan in insular Japonic. Cf. the following data from Blust and Trussel 2020: PMP: *namaw 'sheltered water' , 'deep place in a river' , 'cove' , 'harbor' , 'lagoon' ⟩ WMP: Casiguran Dumagat namaw 'coral reef' , Karo Batak namo 'deep place in a river, generally where it makes a bend and the water moves slowly' , Palauan lǝmáu 'small, deep spot within shallow area inside reef' , CMP: Tetun namo-n 'bay' , 'harbor' , Yamdena name 'deep sea channel in the reef' , Proto-Ambon *nama(w) 'bay' , 'harbor' , Buruese nama 'level stretch in a stream's course; cove, bay, harbor' , POC *namo 'lagoon' ⟩ OC: Wuvulu namo 'lagoon' , Kwaio namo 'lake, pool, deep place in river' , Lau namo 'the lagoon inside a reef, near the reef (the deep) pools towards the shore' , Arosi namo 'a landlocked, shallow lagoon near the shore' , Pohnpeian nahmw 'deep place within the barrier reef' , 'lagoon' , Tongan namo 'lagoon' , Niue namo 'lake' , 'pond' . One has to keep in mind that the above Austronesian words mostly reflect seashore topography. Meanwhile, Kapï is a land-locked province, so semantic shifts after the borrowing are to be expected. Kapï has its good share of lakes, ponds, and rivers, but, of course coral reefs or lagoons are not found there. To illustrate this point, I place below a photograph of Mt. Fuji with some large (unidentified) large body of water in front of it. There are also other makura-kotoba in OJ, which as I suspect, might have Austronesian origin, or be partially Austronesian, but this is still the work in progress that will be presented in a future publication. One of them, is, however, dealt with in the next section of this article. Now I turn to other insular Japonic words that might have Austronesian origin.

Other Austronesian Vocabulary in Insular Japonic
My attention was attracted by the Batanic (Bashiic) languages spoken on a chain of small islands, politically divided between two countries: Taiwan and Philippines. While the Northernmost island, Yami (older name: Botel Tobago, Chin. 蘭嶼 Lán'yú13) 'Orchid Island' , belongs to Taiwan (it is located only 64 km from the southeastern coast of Taiwan, all other islands (Itbayat, Batan,14 and Ibatan), stretching between Northern Luzon and Yami are under Philippines' jurisdiction. In terms of speakers, Ivatan on Batan islands with 35,000 comes the first, followed by Itbayat 3,500, then Yami ca. 3,000, and finally Ibatan (Babuyan) with just 1,240. The current consensus seems to be that the Batanic languages represent a separate WMP subgroup and are not in any case connected genetically to the Philippine languages of Northern Luzon (Ross 2005) and (Blench 2015), however just a few days ago David Zorc and Robert Bust have persuaded me that the Batanic languages are just normal Philippine languages. The Batanic languages are studied comparatively well, with the first publications produced by Japanese scholars, starting with a ground-breaking study of the Yami language by Asai (1936), and also keeping tempo with more recent work by Tsuchida et al. (1987Tsuchida et al. ( , 1989, Yamada (2002Yamada ( , 2014, and Yamada and Zayas (1997). From 1960s Western scholars also pitched in: Reid (1966Reid ( , 1971, Maree (2007), and Maree and Tomas (2012), as well as above-mentioned Ross (2005) and Blench (2015), closely followed by Taiwanese researchers: Li (2000), Zhāng Xún-huì (2000), Yang (2002), and Rau et al. (2012). Finally, there is welcoming sign that Filippino scholars are also getting involved in the study of the Batanic languages: Hidalgo (1998) in addition to the cooperation by Ernesto Constantino in Tsuchida et al. (1989), and C. N. Zayas with Yamada (1997).
Let me start with two Batanic words. The Yami language is naturally saturated with loanwords from Japanese, as the period of Japanese colonial rule in Taiwan was long enough (1895( -1945. Therefore, Japanese loans have penetrated even the Yami basic vocabulary, cf. for example, inu 'dog' ⟨ Jpn. inu 'id.' and saru 'monkey' ⟨ Jpn. saru 'id.' But not all Yami-Japanese look-alikes can be classified in this way. For example, Yami takei 'mountain' (Zhāng 2000: 132) cannot be a loan from Jpn. take (⟨ WOJ takɛ) 'peak' , 'mountain' , or Ryukyuan taki 'id.' , which more frequently appears as u-taki 'sacred grove on a mountain' , 'shrine on a mountain' . There are several facts that strongly speak against the directionality of borrowing from Japanese to Yami in this case. First, Jpn. take 'peak' , 'mountain' is not a colloquial word in the modern Japanese language. 13 Note that this reading is irregular: we should expect Lán xù. 14 Actually, several islands.
It appears mostly either in place names as take ~ -dake, or in poetry. Therefore, it is highly unlikely that soldiers from the Japanese occupational force on Orchid Island would use this word, and not the much more habitual yama 'mountain' in reference to local mountains. Second, Yami takǝi 'mountain'15 looks suspiciously more like PJ *takay rather than its descendant forms. Third, there are apparent cognates at least in the Itbayat and Ivatan languages: Itbayat takey 'field' , 'farm' , 'forest' , 'wilderness' 'island space other than hili (town where people live)' (Yamada 2002: 252); Ivatan takey 'field' (GIEID). Proto-Batanic *takǝi16 was apparently borrowed into insular PJ as *takay due to Arisaka's constraint on the combination of /a/ and /ǝ/ within the same morpheme. Note also that in WOJ takɛ ⟨ PJ *takay functions as a single morpheme and does not have a compounding form taka-resulting from a *-y loss in preconsonantal position. Cf. WOJ takɛ ⟨ PJ *takay 'bamboo' , for which both a free form takɛ and a compounding form taka-are well attested. We can firmly establish the Batanic languages as a source of borrowing into insular Japonic, because as far as I can tell, takǝy 'mountain' , 'wilderness' is not attested anywhere else in the Austronesian language family. WOJ asi 'foot' , 'leg' is an interesting word, appearing only as a free form, the compounding form as well as the form found before the OJ plural suffix -na is just a-, for example ambumi 'stirrup' ⟨ *a-nǝ pum-i 'foot-GEN step-NML' , ato 'footprint' , 'track' ⟨ *a-to 'foot-track' or 'foot-place' , ayupi 'leg cord' ⟨ *a-yupi 'leg-tie-NML' , a-na suwe 'feet end' , etc. Omodaka et al. (1967: 1). The weird behavior of WOJ asi ~ a-'foot' , 'leg' is also paralleled by WOJ unanzi 'nape of the neck' (a free form) and una-'id.' (a compounding form). On the other hand, EOJ a 'foot' appears as a free form in two examples: MYS 14.3387 and MYS 14.3533, but the form asi is not attested in the extant EOJ texts. The EOJ examples are below: (11) MYS 14.338717 (1) 安能於登世受 (2) 由可牟古馬母我 (1) a-nǝ otǝ se-nz-u (2) yuk-am-u koma mǝ ŋ ga (1) foot-GEN sound do-NEG-CONV (2) go-TENT-ATTR stallion (2) [I] want a stallion, which would go (1) without making any noise with its hoofs. 15 Rau and Dong define Yami takei as 'mountain' , 'field' in their online dictionary (2009). 16 In both Yami and Itbayat orthography /ǝ/ is rendered as e. 17 This poem is from Simotupusa province, corresponding to the northern part of modern Chiba prefecture.

Vovin
International Journal of Eurasian Linguistics 3 (2021) 272-300 The tricky part of this example is that EOJ a 'foot' is found before the genitive case marker -nǝ, which always requires a free form of a noun before, cf. EOJ ke-nǝ 'tree-GEN' (MYS 20.4375), unlike WOJ where both free and compounding forms can be found before -nǝ: kï-nǝ 'tree-GEN' (kï is a free form) and kǝ-nǝ 'tree-GEN' (kǝ-is a compounding form).
In Ryukyuan the situation is very different as compared to the one in Japanese. While Ɂasi is attested in Shuri (Kokuritsu kokugo kenkyūjo 1998: 618)18 and asï in Tonosiro (Uchima and Arakaki 2000: 356),19 given the fact that it is a form with -si, which is innovative in Japanese, these two attestations must be viewed as Japanese loans in Ryukyuan. In addition, two other words for 'foot' , 'leg' are attested in Ryukyuan. The first one appears on the periphery as hagi in Yuwan and Koniya dialects on Amami island in the North, and as pagï in Higashi Nakasone, and pagsï in Yonaha on Miyako island, and finally as paŋ in Tonosiro and Hateruma dialects of Yaeyama island group in the South of the Ryukyuan archipelago (Uchima and Arakaki 2000: 356). This word is cognate to WOJ pa ŋ gi 'shin' , MJ fagi 'id.' , and other Japanese forms, such as MdJ -hagi in fukura-hagi 'calf of the leg' , etc. Both WOJ pa ŋ gi 'shin' and PR *panki reflect PJ *panki 'foot' , 'leg' , which underwent a semantic shift in Japanese. There are three independent pieces of evidence that point to PR *panki as an original Japonic word. First, it is preserved on periphery of Ryukyuan, and innovations typically happen in the center, not on the periphery. Second, the semantic shift 18 Shuri attestation is weird, because it appears only in the Japanese to Okinawan index, but not in the main part of the dictionary. 19 Aleksandra Jarosz also informed me that there are other attestations of this word in Miyako, and that it looks like an old loan due to the comound asïda 'geta' and the shift of i ⟩ ï. John kupchik further found anduu 'heel' in Ryukyuan, probably also related to Tōhoku aguto 'id.' The main problems, is however that all these words include opaque 'suffixes' which makes the establishment of a-somewhat not 100% reliable.
'leg' ⟩ 'shin' is easy to imagine, while 'shin' ⟩ 'leg' is not. Third, this semantic shift in Japanese was undoubtedly triggered by the appearance of OJ a 'foot' , 'leg' . Finally, there is Shuri hwisja (Kokuritsu kokugo kenkyūjo 1998: 618), Oku pi̥ sa, Benoki pi̥ sa, Φi̥ sa, Kumejima çi̥ sa, etc. 'foot' , 'leg' (Uchima and Arakaki 2000: 356), found in the center of the Ryukyuan language area. This word is apparently cognate to WOJ pinza 'knee' ⟨ PJ *pinsa 'id.' This time we are dealing with a clear innovation in Ryukyuan: it occurred only in the center, and is a derivation from the word for 'knee' . The important fact is that PJ *panki 'foot' , 'leg' has reflexes in both Japanese and Ryukyuan, while OJ a 'id.' is confined to Japanese. The limited geographical distribution of a 'foot' , 'leg' makes it a perfect candidate for a loanword.
I think that Batanic is again the best candidate for the source of this loanword. I was able to find Yami ai and Itbayat ayi, the latter, according to Blust and Trussel 2020 being the case of a secondary resyllabification in order to confirm to PAN canonical disyllabic structure. Ultimately, Yami ai and Itbayat ayi go back to PAN *qaqay ~ *waqay 'foot' , 'leg' . The situation with ai 'foot' , 'leg' is more complicated than with takǝi 'mountain' discussed above, as it is attested outside Batanic as well. I cite below all cases of ai in Austronesian according to Blust and Trussel 2020. WMP: Bolinaw Sambal a:yi 'foot' , 'leg' , Maguindanao ai 'foot' , 'leg' , Bintulu ai 'foot' , Banggai ai 'foot' , 'leg' , CMP: Tetun ai-n 'leg' , 'foot' , Ujir ai 'leg' , 'foot' , Kamarian ai 'leg' , 'foot' , Wahai ai 'foot' , Hitu ai 'leg' , 'foot' , Batu Merah ai-va 'foot' , Morella ai-ka 'foot' , SHWNG: Pom ai 'foot' . Still, Batanic remains the likeliest source due to its proximity to Ryukyus and Japan. Blust and Trussel posit the following process of resyllabification: *ay ⟩ ai ⟩ ayi. Our task is to determine at which of these stages the borrowing took place. Certainly not at the monosyllabic *ay stage, because *ay would contract to *ɛ in pre-Old Japanese, resulting in OJ e.20 It is the same process that we have already seen above with PJ *takay ⟩ WOJ takɛ 'mountain' , 'peak' . If OJ a 'foot' , 'leg' were from *ay, we would expect WOJ e, not a for its free form. And not at the ayi stage, because it would stay intact in OJ.21 This leaves us with ai 'stage' , which is further supported by OJ phonotactics that prohibits vowel clusters VV. If certain phonological or morphonological processes lead to formation of such a cluster, one of the vowels has to go, and it is normally higher and narrower vowel that does. In the case of Batanic ai, /i/ is certainly higher and narrower than /a/, so in the process of borrowing ai ⟩ a, because ai could not be borrowed as such. As much as I generally dislike etymologies 20 The contrast between PJ *ɛ and *e is lost in OJ in the initial position, after coronals, w and y. 21 See Vovin (2020.1: 48-54) on the contrast /i/ : /yi/ in WOJ.

Vovin
International Journal of Eurasian Linguistics 3 (2021) 272-300 for monophonemic words, I cannot find anything faulty with OJ a 'foot' , 'leg' derived from Batanic ai. The last etymology that I am going to discuss here concerns not the Batanic languages, but the Amis language on Taiwan. As demonstrated by Kupchik (2021), the names of several celestial bodies in Ryukyuan and to a lesser degree in Japanese, come from Austronesian and in particular from Amis. This time, however, we are going to descend from heaven to the depths of the ocean.
There are two words in WOJ for 'whale': kundira and isa. Both words have obscure etymologies, or to put it frankly, no etymologies at all. But there are differences. WOJ kundira is likely to be a non-poetic word: except in the poems KK 9 and its variant NK 7, kundira does not appear even once in whole text of the Man'yōshū, but there are plenty of attestations of isa in the Man'yōshū, while kundira seems to be confined to documents, historical and mythological texts, and dictionaries. Furthermore, kundira undoubtedly enjoys PJ status, as it is everywhere in Japonic: WOJ, MJ, Ryukyuan, MdJ, and modern Japanese dialects. Meanwhile, isa is attested only in WOJ, although there is also a piece of evidence that it was attested as well in the dialect of Iki islands: 俗云鯨爲 伊佐 'In the local dialect 'whale' is called isa' (IFIB 52822). My position has always been that among the two doublets, the one with more limited distribution is a perfect candidate for a loanword.
Indeed, there is Amis Ɂiso and Sakizaya Ɂisu 'whale' . The Amis orthography with o, normally stands for [o] or [ʊ], therefore Wolff believes that it is the same word as PAN *qísu 'shark ' (2010.2: 965). I beg to differ, since sharks are fish, and whales are mammals. Admittedly, people of old might not have such a solid foundation in zoology, as we have today: there are cases when whales are called fish, e.g. WOJ isa-na 'whale-fish' , or Russian folklore рыба-кит 'fish-whale' . Apparently, whatever swims in water is 'fish' . But still such a maritime people as Amis would know the difference between sharks and whales: it would be highly unlikely if both were perceived simply as 'big fish' . Interestingly enough, Blust and Trussel (2020) do not include Amis Ɂiso 'whale' under their PAN *qiSu 'shark' , the only representative of Formosan languages cited is Paiwan qisu 'shark' . There is no dearth of numerous WMP and CMP forms provided in Blust and Trussel 2020, but they are not relevant to our purposes here, as none of them could be a source for WOJ isa.
Thus, at the initial glance, Wolff's Austronesian etymology for Amis Ɂiso 'whale' has a semantic problem, while WOJ etymology isa 'whale' has a phonetic problem of vowel correspondences Amis /u/ : WOJ /a/ in the second syllable. I will attempt to demonstrate now that this problem with different vocalism 22 Pages are cited according to Akimoto Kichirō's edition of Fudoki.
in Amis and WOJ is actually a pseudo-problem. First, the phonetic sequence [isu] within the same morpheme in WOJ is extremely rare. To the best of my knowledge, it is found only in five cases: 1) Isu-nǝ kami 'Top of the rock' , a name of a famous shrine, where isu is certainly a result of *o ⟩ u raising, and Iso-nǝ kami, an original form that is attested wider and better than Isu-nǝ kami; 2) in three WOJ words: isusuki, isusuku, and isurǝkǝp-with unclear meaning; 3) in isu kupasi, a makura-kotoba for kundira 'whale' , which is also considered to have an unknown meaning (Omodaka et al. 1967: 75-76), Actually, this makurakotoba is extremely important for our purposes here. I think that it represents an Austronesian-Japonic hybrid, where isu is Austronesian, being a loan from Amis Ɂiso 'whale' . The remaining part is Japonic. Certainly, it is unlikely that we are dealing here with WOJ kupasi 'to be beautiful' , 'to be perfect' , 'to be superb': 'whale (kundira) that is beautiful [like] whale (isu)' would make little, if any sense at all. The primary function of whales in Japan from times immemorial to the present day, has always been a food source. Therefore, I think that WOJ kupasi in this case consists of kup-'to eat' and adjectivizer -asi, invariably found after verbal stems. If I am right, then isu kup-asi may not seem as syntactically modifying kundira, because the attributive form of -asi is -asi-ki. However, since we are dealing with poetry, -ki could have been also omitted to confirm to 5-7 poetic meter. Also, attributive in -ki in WOJ is a relatively recent innovation, and as Kupchik indicated to me, it is unlikely to predate this makura-kotoba. Then isu kup-asi kundira can be understood as 'a whale, which is an edible whale' . It is also interesting that all attestations of isa in the Man'yōshū are in the compound isa-na 'whale-fish' ,23 and all of them are found invariably in the context referring to whale hunting like in the example from MYS 17.3893 below. Therefore, it seems that we have both isu and isa 'whale' in WOJ. I suspect that the second syllable vowel a in isa is due to the regressive assimilation that took place in the compound isa-na ⟨ *isu-na, as well as to overall structural pressure to eliminate the sequence [isu]. Relevant WOJ textual examples (13-15): (13) KK 9 伊須久波斯久治良佐夜流 isu kup-asi kundira sayar-u whale eat-ADJ whale be.caught-FIN A whale, an edible whale, got caught.

Austronesian Affixes in Insular Japonic
These are a priori difficult to find, because Austronesian uses mostly prefixes and infixes, while in Japonic suffixes rule supreme, except in OJ, which has some prefixes as well. Thus, for example, there is an OJ locative prefix sa-, which marks not the separate nouns, but NP. This prefix is not recognized in the traditional Japanese grammar, but see its full description in Vovin (2020.1: 101-105, 108-109) As noted in Vovin (2020.1: 109): "There is a dative-locative case marker -sa in all Tōhoku dialects. The proper locative function seems to be confined to Tōhoku (those dialects that have it do not have Central Japanese -ni). In directive and dative functions case marker -sa is found also in Northern and Southern Kyūshū and in Northern Kantō". In peripheral modern Japanese dialects (Tōhoku and Kyūshū), where it probably became a suffix under the pressure of the SOV word order.
More than fifteen years ago, in the first edition of my WOJ grammar I have suggested that this sa-might have an Austronesian origin (Vovin 2005: 90-91), repeated in Vovin (2020.1: 109). At this time, I mentioned the existence of the locative sa-in Tagalog, Atayal, and Rukai, noting also that in Atayal sa-marks NP like in OJ: The same holds true in Bikol (sa locative class marker occurring before general nouns, see Mintz 1971: 368) and Maranao (san 'there' , 'in' , 'at' , see McKaughan and Macaraya 1967: 389)

. No textual examples for any of them.
This list is probably far from conclusive, but accessible resources on the Philippine languages that I am able to reach are quite limited during the COVID-19 pandemic years. But even on the basis of the short list of attestations above, one can see that the locative sa is found in the Northern, Central and Southern Philippines. This is important, since both Atayal and Rukai are the languages spoken in the interior of Taiwan. Consequently, their speakers are not seafarers, and the Austronesian source of OJ sa-is unlikely to be in Taiwan. It is a different story with Philippines, which seem to be the likeliest source of OJ sa-.
In addition, OJ directive-locative focus verbal prefix i-might also have had an Austronesian origin. The locative prefix i-attested in the Amis language is, on the contrary, a nominal prefix. This fact makes this etymology less credible, but not impossible. Furthermore, given the situation that the Amis language as demonstrated by Kupchik (2021), is the most prodigious source of Austronesian loans in the insular Japonic, I think that it is even more or less likely. Amis (Cài and Zēng 1997: 307) a. adihay ko kapaysinan no ' Amis i-tiya ho be.many SM taboo GEN ' Amis LOC-long.time ago There were many taboos for Amis in the times now past.

26
The following examples are all from (Cài and Zēng 1997). However, since Cài and Zēng's glosses are sometimes incomplete and also include omissions, I used the Amis dictionary by Virginia Fey (1986) to fill in the necessary gaps.
(Cài and Zēng 1997: 306) b. i-tiraay ingata no lalan no tosiya ko niyaro' LOC-there be.close GEN road GEN car SM village The village there is close to the motor-way.
(Cài and Zēng 1997: 306) c. hinam-han nira i-taliyok no talo'an ira ko mapeliay a narakatan no tamdaw look.around-DEB he LOC-perimeter GEN small.hut exist SM be.left CP trace GEN person When he was prompted to look around the small hut, there were traces left by a person.

Conclusion: When and Where?
I will try to give very approximate and tentative answers to two questions here: "When these Austronesian loanwords entered insular Japonic?" and "Where did it happen?" The first question is easier than the second. I do not see any traces of these loanwords in peninsular Japonic, therefore the upper temporal boundary must not be earlier than 300 BC, the rough date when the Yayoi culture that could be defined as Proto-Japonic first appeared in Japan. Although it has become trendy recently to push back the beginnings of the Yayoi culture to 700 BC or so, I am not convinced, because the evidence presented is not very reliable. As for lower temporary boundary, 400 AD seems to be quite likely, because while our knowledge of Ancient Japan increases exponentially after this date, leaving no likely room for significant population movement or language contacts on a large scale (and judging by the nature of these loans, the language contact was indeed on a very large scale, since we are dealing not with just a peripheral, but with a basic vocabulary). However, before 400 AD we know precious little about the ethnolinguistic history of Japan.
The second question is much more difficult to answer. However, at the present point we already have several pieces of the puzzle at our disposal. First, we know that the initial landfall of the Yayoi culture in Japan was in Northern Kyūshū, from where it steadily and rapidly spread to the rest of Kyūshū, Shikoku, as well as to the most of Honshū, excluding its Northern part, which was dominated by Ainu at least until the late ninth or early tenth century AD. Ainu-Japanese bilingualism was still present in the Kantō region of Japan well into the early tenth century AD (Vovin 2009b(Vovin , 2022. Second, we know Vovin International Journal of Eurasian Linguistics 3 (2021) 272-300 that Proto-Ryukyuan speakers did not start to migrate to Ryukyuan islands earlier than the late ninth century or early tenth century AD. There are different opinions as to where Proto-Ryukyuan was spoken: in the North-Eastern Kyūshū, North-Eastern Kyūshū, or Southern Kyūshū (Serafim 2003), but we can say for sure that the Austronesian loanwords detected by Kupchik (2021) in Old Okinawan come back to the time when Proto-Ryukyuan was still spoken on Kyūshū. Third, although we know very little about the Hayato (WOJ Payatǝ, 隼人) and almost nothing about Kumaso (WOJ Kumasǝ, 熊襲) tribal languages in Kyūshū, their customs, headgear, clothes, shields, etc. look strikingly Austronesian (Ōbayashi 1975: 24-33). A treatment of the tits and bits of the Hayato and the Kumaso languages would require a separate article, so here I will limit myself to just one note. Almost half a century ago it was suggested by Murayama that WOJ -sǝ in Kumasǝ and -tǝ in Payatǝ reflect PAN *Cau 'person' , 'human being ' (1975: 258-260). Apparently, there is a phonetic difficulty: we would rather expect WOJ -to and -so rather than WOJ -tǝ and -sǝ as a reflex of PAN *Cau. But there is a spelling Kumasǝo (球磨贈於) in Fudoki (CFIB 503), (HFIB 520, 521), (BF 358, 368), (HF 380), or almost identical 球磨囎唹 (HF 404). This spelling apparently violates the OJ phonotactics prohibiting vowel clusters, and, consequently could only be an attempt to represent a foreign word. It still does not explain the vowel [ǝ], but, nevertheless brings us one step closer to PAN *Cau. The Extra-Formosan MP languages all uniformly have initial consonant t-, while the Formosan languages show here a variety of reflexes: Papora sō, 'person' , 'human being' , Pazeh saw 'person' , 'human being' , Thao caw 'person' , 'human being' , Hoanya sau 'person' , Tsou cou 'person' , 'human being' , Puyuma Tau 'to be human' (Blust and Trussel 2020). For WOJ -sǝ Formosan origin will work better, but for -tǝ either Formosan or MP will do, because OJ has never had any affricates [c] or [č]. A possible solution to [ǝ] vowel may be offered by Tagalog that has táɁo 'person' , 'human being' with penultimate stress, but taɁó 'mortal' with a stress on the last syllable. It is quite possible that WOJ sǝo (贈 於, 囎唹) represents an attempt to transcribe sǝɁo. Furthermore, unfortunately our knowledge about OJ accentual system is speculative at best, but if sǝɁo had an accent pattern LH, it is not inconceivable that the vowel [ǝ] is reduction of an original *a. Therefore, we can posit a hypothetical *saɁu, which will be quite in agreement with Austronesian forms.
All three independent pieces of evidence discussed above point in the same direction: the quite intensive language contact between Austronesian and insular Japonic took place on Kyūshū sometimes between 300 BC and 400 AD. There will be many years ahead of the hard work before we get more or less coherent general picture, but it looks like that starting with Kupchik (2021) a new subfield of Japonic historical linguistics is born.
To put an icing on the cake, it is quite possible that Austronesians might have travelled to much more distant lands in the North than Japan. Long ago I have noticed PE *iqaluγ 'fish' (no Aleut cognates). In this word *-luγ looks like a postbase, but it cannot be -luγ 'to be bad' , and needs to be explained. The remaining part *iqa-suspiciously looks like PEMP *ikan 'fish' ⟨ PMP *hikan ⟨ PAN *Sikan. Aleut tana-ẋ 'land' , 'earth' (traditionally compared with Eskimo nuna 'id.' , but this etymology has too many problems) no less suspiciously looks like Maranao, Manobo, Visayan, etc. tanaɁ, Malay, Old Javanese, Javanese, Balinese, etc. tanah ⟨ PMP *taneq 'land' , 'earth' , 'soil' (Blust and Trussel 2020). Certainly, these might be just chance look-alikes, but, in my opinion, this possibility is worth investigating a little bit further, in spite of the fact that poor Austronesians would be chilled to their bones in these frigid Arctic waters.