Northeast Asian Linguistic Ecology and the Advent of Rice Agriculture in Korea and Japan

The Shandong/Liaodong dispersion hypothesis

Miyamoto (2009) presents a scenario where agriculture spread from (possibly distinct) locations in northeast China to the east and north (Fig. 2). Miyamoto distinguishes three major spreads of agriculture from this region. The first, associated with Setaria and Panicum, spreads from the Liaoning region to the northwestern part of the Korean peninsula and thence to its eastern and southern coasts. Almost simultaneously, a spread of these cultivars took place from northeast China to the south of Primoriye and thence to its coastal plain (Miyamoto 2009: 25–6). Miyamoto dates this first spread to the first half of the fourth millennium BCE. The second dispersion adds dry field rice cultivation to the millets; it spreads from Shandong through the Liaodong peninsula to the south of the Korean peninsula in the second half of the third millennium BCE. The third dispersion includes wet rice cultivation and associated tool complexes. It takes the same route from Shandong to Liaodong to the Korean peninsula around the middle of the second millennium BCE. In the Korean peninsula, it is associated with the Mumun (plain, patternless) ceramic culture and irrigated paddy cultivation. This third dispersion reaches the Japanese archipelago, where wet rice cultivation appears in northern Kyūshū around 950 BCE (800 BCE according to Miyamoto 2009: 28).²

Open image in new window

Miyamoto’s scenario for the third dispersion is consistent with the scenario for the introduction of wet rice cultivation into the Korean peninsula provided by Ahn (2010). Ahn discusses and dismisses claims for very early rice cultivation in the Korean peninsula and finds inconclusive evidence for rice cultivation during the Chŭlmun (comb pattern) period (ca. 6000–1300 BCE). Previous researchers (e.g., Crawford and Lee 2003) concur on the evidence for Chŭlmun cultivation of Setaria and Panicum diffused from the Liaoning region, consistent with the Shandong/Liaodong dispersion hypothesis. Ahn argues against earlier views that wet rice cultivation was introduced into the Korean peninsula from the south, and concludes in favor of the scenario where it enters the peninsula from Liaodong in the early Mumun period around or after 1500 BCE. Thereafter, evidence for agricultural settlements engaged in rice farming is found throughout the peninsula with the exception of the northeast. The relative importance of wet rice cultivation shows regional variation throughout the Mumun period: wet rice is dominant in the central and southwest regions, while millet and other dry field crops are relatively more important in the southeast (Ahn 2010: 93).

Hiatus in wet rice agriculture on the Korean peninsula

According to Ahn (2003: 81–81; 2010: 91), rice farming settlements disappear from the archaeological record during the third century BCE, reappearing again in the first century CE. Ahn suggests that the specialized nature of intensive rice paddy farming might have been especially vulnerable to environmental fluctuations, but he acknowledges that there is no evidence for major climatic changes that might have triggered a hiatus in rice farming during this period (Ahn 2010: 97). The hiatus in wet rice sites coincides with the emergence of the Korean-style bronze dagger culture in the third century BCE. As described by Ahn (2003), this culture spreads from the Kŭm river basin on the west coast of Korea. In addition to its characteristic bronze daggers, this culture is characterized by coarse patterned bronze mirrors with multiple attachment loops (多鈕粗文鏡), shield- and hilt-shaped bronze implements, black long-necked earthenware, and clay-rimmed ceramics. Ahn emphasizes that the ceramic style, pattern of settlement, and burial styles of this culture indicate a break with the previous Mumun culture, although the two cultures continue to coexist for some time. The Korean-style bronze dagger itself descends from the Liaoning bronze dagger prototype. The Liaoning prototype is found in the Korean peninsula from 1300 BCE on and in Kyūshū from 800 BCE, but Ahn (2003) dismisses the possibility that the distinctive Korean-style bronze dagger developed from this prototype within the Korean peninusla. Instead, he traces the origins of the Korean bronze dagger culture to the central Liaoning region, where similar pottery, burial, and dwelling styles are found in association with the Liaoning bronze dagger prototype. According to Ahn, this Liaoning culture split into two branches, one of which remained in Liaodong and the Changbaishan region. The second branch brought the Korean-style bronze dagger culture to the Kŭm river area. Ahn writes of this culture: “Agricultural settlements disappeared from the archaeological record from the third century B.C. when the Late Mumun culture,³ with a nomadic lifestyle, spread from the Liaoning region of northeast China” (Ahn 2010: 91). It is not clear to what extent the Korean-style bronze dagger culture should be characterized as nomadic, but it is clear that it was not, in its initial appearance, associated with wet rice cultivation. The advent of this culture can account for the temporary disappearance of wet rice farming settlements between the third century BCE–first century CE.

Early Chinese sources on the Korean peninsula and the Japanese archipelago

The earliest substantial Chinese sources on populations in the central/south Korean peninsula, the so-called Dongyi “Eastern barbarian” (Book of Wei) in the Wei shu section of the Sanguo zhi (late third century CE) and the Hou Han shu (fifth century CE), describe three groupings of peoples, the so-called Samhan “Three Han” (三韓 Sanhan): Mahan (馬韓) in the west central region, Chinhan (辰韓 Chenhan) in the southeast, and Pyŏnhan (弁韓 Bianhan) in the south.⁴ Traditional Korean historiography indentifies these groupings as the antecedents of the historical polities Paekche (百済), Silla (新羅), and Kaya (加耶), respectively. Inscriptional evidence indicates that the term Han 韓 had ethnonymic significance not only for the Chinese authors of the Wei shu and Hou Han shu but for local Sinoxenic peoples as well. Thus, the inventory of gravekeeper villages on the 414 CE stele memorializing Koguryŏ king Kwangaet’o (廣開土) lists both Koguryŏ 高句麗 and Han 韓 villages, without distinguishing the latter as Mahan, Chinhan, etc. Nevertheless, the Wei shu and Hou Han shu present a picture of some ethnic and linguistic diversity in the Samhan region in the third century.

The Wei shu describes the language of Chinhan as “not the same as that of Mahan” (其言語不與馬韓同).⁵ While the people of Mahan are described as not knowing how to use cows and horses as means of transport (不知乗牛馬), those of Chinhan use them. The latter are described as planting the five crops with the addition of rice (宜種五穀及稻), suggesting that the latter may have been something of an add-on.

The Wei shu describes the Pyŏnhan and Chinhan populations as “living intermingled together” (弁辰與辰韓雑居). It describes their clothing and dwellings as the same, and their languages and customs as similar (言語法俗相似). The Hou Han shu begins with the same phrase about intermingled living, then states “enclosed towns and clothing are all the same” (城郭衣服皆同) but “languages and customs have differences” (言語風俗有異). This section of the Hou Han shu, as the later of the two texts, often cites the Wei shu, but the Hou Han shu drew on other sources as well. The different descriptions of linguistic and cultural distinctions may reflect differing responses to a situation of ethnic complexity. Similar complexity is suggested by the descriptions of physical type. The Wei shu states that “Chinhan men and women are close to Wa (男女近倭),” the ethyonym for the contemporary inhabitants of the Japanese archipelago, and like the Wa tattoo their bodies. The Hou Han shu identifies this as a feature of Pyŏnhan, stating that “their country is close to Wa, therefore they frequently have tattoos.”

Linguistic differences are confirmed by the Wei shu toponyms for the Samhan. The Wei shu gives phonogrammatic spellings for 54 Mahan settlement names. As pointed out by Toh (2008: 234–5), most are disyllabic: 34 are transcribed with two syllables, 10 more with two syllables and a suffix, one of which is identifiable as *-pieliai 卑離, usually related to the Paekche word puri <夫里> “town,” itself typically compared to Late Middle Korean -βɨr “town.”⁶ These suffixes do not occur in the Chinhan and Pyŏnhan settlement names. The Wei shu lists the latter together in no particular order. It prefixes the Pyŏnhan names with Pyŏnjin 弁辰. However, the Chinhan and Pyŏnhan toponyms also appear to draw on two distinct linguistic traditions. All but one of the 12 Chinhan names are disyllabic (like the Mahan names). Five of the 12 Pyŏnhan names have three or more syllables, and three of these appear to involve a suffix, *-mietoŋ 彌凍and *-jamaʔ 邪馬. The first of these suffixes also occurs in the only Chinhan polysyllabic settlement name.⁷ The second is identical to the transcription given in the Wei zhi for the first two syllables for the contemporary *Jamaʔdə “Yamato” grouping in the Japanese archipelago. Scholars generally intepret these as involving a morpheme cognate with proto-Japonic *jama “mountain” (Bentley 2008: 14). A virtually identical spelling *jama 邪麻 occurs as an independent word in the Wa toponyms on the Japanese archipelago given in the Wei shu.

In light of the discussion in the “Hiatus in wet rice agriculture on the Korean peninsula” section, a simplistic interpretation of the Wei shu and Hou Han shu descriptions might be that the three Han groupings correspond to three distinct but related ethnicities. In fact, the texts indicate a more complex (and plausible) interrelationship between language, ethnicity, and protopolitical grouping. Mahan, the larger, better established grouping, occupies the area where the Korean-style bronze dagger culture emerged some five centuries earlier. Chinhan represents a population more recently arrived from the northwest, as indicated by its oral traditions and its mastery of animal husbandry. The Chinhan population lives intermixed with Pyŏnhan; the Chinese reporters struggle to decribe the resultant demographic complexity. Their languages may be similar, or different; some resemble the Wa, some tattoo their bodies. While Wa-like toponyms are more frequent in the Pyŏnhan grouping, one such toponym is identified with Chinhan. This is exactly the kind of complexity we might expect to be associated with the situation described by Ahn, where a population associated with Mumun wet rice growing culture lives alongside more recently arrived members of the Korean-style bronze dagger culture.

Toponymic evidence on language locations

There are two broad interpretations of the Koguryŏ phonogrammatic material. One takes them to represent the Koguryŏ language. This interpretation is adopted in earlier Japanese scholarship, by the Korean scholar Lee Ki-moon (see Lee and Ramsey 2011) and by Christopher Beckwith (see Beckwith 2007). It has been influential among anthropologists, e.g., Hudson (1999). The second interpretation, associated with the Japanese scholar Kōno Rokurō and the Korean linguist Kim Bang-han (Kim 1983), claims that the Koguryŏ phonogrammatic material transcribes the toponyms of linguistically distinct, non-Koguryŏ peoples.

Scholars adopting the first view have arrived at diametrically opposed conclusions about the nature of the Koguryŏ language. Thus, Lee and Ramsey (2011) emphasize the lexical material in the Koguryŏ language relatable to Korean,¹⁰ while Beckwith (2007) considers Koguryŏ to be a continental relative of Japanese. In contrast, the second view explains why the Koguryŏ phonogrammatic material transcribes words relatable to Japonic and words related to Koreanic. Koguryŏ used phonograms to transcribe indigenous names from languages other than their own. They also devised standard Chinese binomic names for some localities that came under their control; for such localities, the two names coexisted.

From the standpoint of this paper, the important takeaway lesson from the Koguryŏ toponymic data is that a language cognate to Japonic was spoken on the Korean peninsula. This is a point of consensus for all major scholars who have worked on this material. The range of the Koguryŏ toponymns is confined to the region of historical Koguryŏ control, so they provide no information about the southern tip of the peninsula, but the northern range of phonogrammatic toponyms with widely accepted Japonic interpretations extends as far as modern North Hwanghae province, south of the later Koguryŏ capital at P’yŏngyang.

Chronological depth of language families

Both Japonic and Koreanic are relatively shallow language families. Comparative phonological evidence shows proto-Japonic to be somewhat older that the oldest extensive textual attestations of Western Old Japanese in the eighth century. Proto-Ryūkyūan maintains the distinction between proto-Japonic *e and *i and *o and *u in wider range of environments than does Western Old Japanese, indicating that the ancestor of pR diverged from a parent older than WOJ (Hattori 1977–1979). Phonological information like this provides a ceiling but not a floor for the date of the protofamily; however, a radically earlier date would lead us to expect a greater degree of phonological divergence. Hattori’s (1953) glottochronological study estimates a date of 500 CE for the divergence of the ancestors of Early Middle Japanese and Shuri Ryūkyūan. Hattori arrives at this date by adjusting the logarithmic decay function proposed by Swadesh to fit the facts of several known cases of divergence.

A standard criticism of glottochronology is that it assumes a constant rate of vocabulary replacement across languages. In a recent paper, Lee and Hasegawa (2011) attempt to overcome this and other defects of glottochronological approaches using a Bayesian phylogenetic analysis based on lexical data from 59 Japonic varieties. This model assumes a single rate of vocabulary substitution across varieties, but the rate is calibrated on the basis of known historical dates (in this case, those for Western Old Japanese and Early Middle Japanese). The phylogeny selected by Hasegawa and Lee is problematic in its shallower branches, which represent all non-Ryūkyūan branches as descended from EMJ, but it is not clear that this affects their overall results. Lee and Hasegawa estimate a date of 2182 BP for the ancestor of proto-Japonic. This result is important because it disconfirms the possibility of Kofun period (third to sixth century CE) date for pJ, something not completely disallowed by Hattori’s results. However, a Kofun period date for pJ would also be inconsistent with the toponymic evidence for Japonic on the Korean peninsula, unless the toponyms somehow resulted from a later historical movement of Japonic speakers to the continent.

Phonological evidence indicates that proto-Koreanic is even shallower than proto-Japanese. The evidence is similar: data from the Cheju variety show a broader distribution of the back central unrounded vowel/ʌ/than is found in fifteenth century Late Middle Korean texts. Once again, this gives us a ceiling for divergence somewhat earlier than the fifteenth century; once again, if the protolanguage was radically older, we might expect greater phonological divergence.

Neither the phonological evidence nor the statistical evidence (in the case of Japanese) is consistent with a date of protolanguage divergence older than the dates for the beginning of wet rice agriculture, as pointed out by Hudson (1999) and Lee and Hasegawa (2011). This fact alone does not rule out the possibility that proto-Japanese descends from a pre-Yayoi Jōmon language, or proto-Korean from a pre-Mumun Chulmun language. In either case, it is a prima facie possibility that such a language, indigenous to the region prior to the arrival of wet rice agriculture, expanded and replaced previously existing indigenous languages as a result of the demographic expansion associated with the new agricultural technology. In the case of Japonic, however, once again, such a scenario would have a difficult time explaining the toponymic evidence for Japonic on the Korean peninsula.

The scenarios whereby Japonic arrived in the archipelago and dispersed as a result of the Yayoi expansion, and Koreanic arrived in the peninsula and dispersed as a result of the advent of the Korean bronze curved dagger culture, are consistent with the farming/language dispersal model (Bellwood and Renfrew 2002). The dates of these two events, 950 BCE and 300 BCE, respectively, are also consistent with the gap between the chronological ceilings for dispersal of the two families, before 700 CE for Japonic and before 1,450 CE for Koreanic. In both instances, we know that the actual date of dispersal must be earlier, but we do not know how much. Even Lee and Hasegawa’s date, first century BCE, leaves a 900-year lag between the archaeological event and the linguistic evidence for dispersal.

The remarkable non-diversity of Japonic and Koreanic can be explained by two factors. The first is a founder’s effect, the phenomenon by which genetic diversity is reduced when a small population settles a new area. The claim that a serial founder’s effect is discernible in linguistic variation has been made by Atkinson (2011), among others. In the case of Japonic, we might expect founder’s effects to have occurred as a result of the movement of relatively small populations from the Korean peninsula to Kyūshū, and again from Kyūshū to the rest of the archipelago. Crudely put, the effect can be conceptualized as a local reduction in linguistic diversity compared to the home population. The same effect would be anticipated in the establishment of Koreanic in the south central peninsula, and again as it expanded throughout the peninsula.

The second factor is archaeohistorical. Both the Yayoi expansion in Japan and the spread of the Korean bronze curved dagger culture were subject to bottleneck effects, to borrow another term from evolutionary science. Kobayashi (2007) accounts for the relatively slow spread of Yayoi culture to the east in terms of the “walls” (壁 kabe) put up by the progressively more robust Jōmon cultures to the east in the archipelago. In the case of Korea, the Chinese commanderies in the north of the peninsula imposed a bottleneck until their demise in the early fourth century CE. Release of each bottleneck results in a new dispersal and founders effect. These effects leave phonological traces; thus, the categories of lexical accent are less complex in eastern Japonic varieties, while Koreanic varieties in regions to the north and east, as well as Cheju, lack lexical pitch accent altogether.

Note that the founders effect scenario presupposes demic diffusion. Thus, the relative nondiversity of Koreanic and Japonic provides further support for the view that the speakers of the protolanguages arrived from elsewhere.

Rice and related agricultural vocabularies

Vovin suggests cognates for four of these, (2e), (2g), (2i), and (2j) from Koreanic, with cognates for (2e) in Tungusic and Turkic, for (2g) in Mongolic and Turkic, and for (2i) in Tungusic as well. He finds no external etymologies for (2b), (2c), and (2h). Vovin specifically rejects Austronesian cognates proposed in earlier research, but he suggests Austroasiatic cognates for (2a), (2d), and (2f). Sagart (this issue) has proposed an alternative Sino-Tibetan-Austronesian etymology for (2d).

As this discussion suggests, it is not a straightforward matter to identify cognates in this lexical domain in Northeast Asia, and it is not straightforward to distinguish inherited cognates from loans. In this section, I will confine myself to some general observations about rice-related vocabularies in Korean and Japanese and possible relations between them as they relate to the Shandong–Liaodong dispersal hypothesis.

As Vovin observes, some of the items in (2) are the products of internal semantic specialization, such as (2b) *mə/omi “unhulled rice” < *mə/om- “pound” + *-i nominalizer, and (2e) *ipi “cooked rice,” which Vovin derives in a similar way from a verb *ip- “eat.” These derivations raise the possibility that the ancestor language lacked specialized terms for “unhulled rice” and “cooked rice.” Similarly, (2f) and (2i) are not specialized rice-related terms. The lack of specialized vocabulary specifically dedicated to rice is more visible in Korean, where the terms in the semantic role of (2c-d) psʌr H < *pʌsʌr and (2e) pap L both designate hulled and cooked grains, respectively, of any type. The Korean (and Altaic) cognates that Vovin suggests for (2e), (2g), (2i), and (2j) are all unspecialized: they mean “eat,” “field, plain,” “flour, powder,” and “malt.” Corresponding to (2), the only semantic category with a Korean term specialized for rice is (2a) Late Middle Korean pjə H “riceplant.”

These facts are consistent with the Shandong/Liaodong dispersion hypothesis outlined in “The Shandong/Liaodong dispersion hypothesis” section. If the language families commonly grouped together as Altaic are related to Japonic, they presumably dispersed from Shandong prior to Japonic since their historical ranges are more remote. Even in the case of Koreanic, if Ahn’s hypothesis that the Korean-style bronze dagger culture entered the peninsula from central Liaoning is correct, Koreanic may represent an earlier, pre-rice cultivation dispersion from Shandong. Any rice cultivators left behind in the greater Shandong region after Miyamoto’s third dispersion were absorbed by the expansion of Sinitic, so no trace of their languages remain there. Cognate vocabulary between the surviving languages dispersed from Shandong precedes rice cultivation.

Summing up the results of this section, Japonic gives some evidence for agricultural vocabulary cognate with other languages in Northeaast Asia, but none of this vocabulary is dedicated to rice. Koreanic shows relatively little vocabulary dedicated to rice at all. These facts are consistent with a dispersal of some languages, including Koreanic, from Shandong prior to the advent of wet field rice cultivation in that area. The cognate agricultural vocabulary shared by Koreanic and Japonic precedes wet rice agriculture.