On the Japanese homeland

Bayesian phylogeny is becoming a popular method to reconstruct the structure and potentially the homeland of various language families. For example, a study conducted by Quentin D. Atkinson and discussed in an earlier posting uses this method to determine the timing and the geographical location of the Indo-Europeans' homeland. Another recent study, by Sean Lee and Toshikazu Hasegawa of the University of Tokyo, applies the same methodology to the question of the Japonic homeland.

Linguistically speaking, the Japonic language family includes not only Japanese proper (spoken by 121 million people in Japan) but also a number of smaller Ryukyuan languages, such as Okinawan (984,000 speakers) and Kunigami (5,000 speakers) spoken in Central Okinawa. The question of whether Ryukyuan varieties are separate languages or dialects of Japanese remains open; however, treating Japonic as a language family rather than a single Japanese languages becomes more and more popular.

Archaeologists have long known that the Japonic speakers, that is the Japanese and Ryukyuan peoples, grew out of two ancient groups, known as the Jōmon culture and the Yayoi culture. The term jōmon refers to a straw-rope pattern, created by using straw-ropes to decorate pottery. The Jōmon group started out as nomads but gradually switched to a sedentary lifestyle. Rather unusually, the Jōmon remained hunter-gatherers; it is believed that the Japanese islands were so abundant in food, which, combined with the early knowledge of pottery allowed the Jōmon people to live relatively stationary lives without farming.

However, around 400 BCE there seems to have been a great change which led the people to shift from hunting and gathering to farming, to adopt irrigated rice paddies as well as weaving, bronze tools and ironware. All this led to a dramatic increase in population. This new culture is known under the term Yayoi, but little is known on how these novel cultural and technological traits came to Japan: invasion, immigration or cultural transmission.

One way or another, the Jōmon and Yayoi people probably interbred, resulting in a mixed population from which descend the modern Japanese. In particular, genetic studies suggest that the ancient Jōmon people contributed as much as 40% of their DNA to the modern Japanese gene pool.

However, when it comes to language, we never say that a given language is "40% of family X and 60% of family Y". For example, despite heavy influences of Latin and (Norman) French, English remains a Germanic and not a Romance language. Similarly, despite much borrowing from Turkic languages, Russian remains a Slavic language. So, are modern Japonic languages descendants of the language of the earlier Jōmon people or of the language of the relative newcomers, the Yayoi people? And what other languages or language families are these languages of the ancient groups related to?

The study by Lee and Hasegawa relied on Bayesian phylogeny, the method depending on having a computer generate a large number of possible family trees and sampling them to find the most probable one. The tree selected by this method can be dated, provided any fork in this tree can be linked to a historical event. In this study, the dates for Old Japanese, Middle Japanese and the split between the Kyoto and Tokyo dialects (which began in 1603 CE, when the Japanese capital was moved from Kyoto to Edo, the then-name for Tokyo) were used to date the Japonic tree. The date that was arrived at for the root of the Japonic tree, 2,182 BP ("before present", or 171 BCE) correlates well with one dating for the Yayoi culture, suggesting that it was the Yayoi people who brought the ancestor of the Japanese language to Japan (much like the Anglo-Saxons brought the ancestor of Modern English to the British Isles). However, a new dating of the Yayoi based on a recalibration of radiocarbon dates pushes them to as far back as 3,000 BP, or 800 years earlier than the root of the Japonic tree, according to Lee and Hasegawa.

Furthermore, even if it is correct that the Yayoi brought the ancestral Japonic language to Japan, the question remains as to whether they came from the Korean peninsula (as Lee and Hasegawa seem to assume) or from elsewhere, for instance, from Taiwan. Archaeologists remain divided on this issue: some (cf. Ottosson and Ekholm 2007) think that the Yayoi culture spread to Japan from the Korean peninsula, while others (cf. Barnes 1999) trace them to Taiwan. Among linguists the trend is to associate the Jōmon people with the Austronesian family of languages, whose homeland was most likely in Taiwan or in the adjacent areas of coastal mainland China, and the advent of the Yayoi culture with Altaic-speaking groups. This led scholars like Ono (1970) to propose a mixed Austronesian-Altaic hypothesis for the origin of the Japanese language.

Note, however, that most of the evidence for either the Altaic or the Austronesian connection of Japanese comes from the examination of the lexicon. Essentially, scholars look for probable (to them, all but proven) cognates among the words in the so-called Swadesh list. This list, named after Morris Swadesh who proposed in 1955, includes 100 words that are said to constitute the core vocabulary (presumably, the most conservative part of any language's lexicon).

According to Robbeets (2005), 45 words among the 100 words in the Swadesh list are Altaic cognates. Included in this list of 45 Altaic cognates are words for pronouns 'I' and 'you', numerals 'one' and 'two', adjectives like 'big', 'small', 'red', 'white' and 'new', body parts like 'blood', 'hand' and 'heart', other nouns like 'man', 'sun' and 'fire' and verbs like 'eat', 'sleep' and 'kill'.

Another list of probable cognates was composed by Benedict (1990), but this is a list of probable Austronesian cognates. It too includes 45 out of 100 words in the Swadesh list. Based on these figures alone, there is a draw between the two hypotheses for the origins of the Japanese language: the Altaic or the Austronesian. However, the situation is more complicated that that: even though the Altaic and Austronesian hypothesis each claim 45 cognates from the Swadesh 100-word list, 21 of these cognates overlap! This list of overlapping cognates includes words for 'I', 'one', 'two', 'small', 'tree', 'skin', 'blood', 'bone', 'mouth', 'tooth', 'foot', 'hand', 'belly', 'eat', 'bite', 'sun', 'star', 'earth', 'fire', 'yellow', 'round'.

Obviously, both Robbeets and Benedict cannot be right: a word cannot trace its origin to both Altaic and Austronesian simultaneously (unless we can show independently that those two families are closely related, which we have no evidence for whatsoever). This means that either Robbeets (2005) with her list of Altaic cognates or Benedict (1990) with his list of Austronesian cognates is wrong, or most likely both of them are partially wrong in claiming too many cognates for their favorite hypothesis.

By the way, one of the main difficulties in establishing likely cognates is the tendency for consonant-vowel (CV) syllables: given this restriction on possible syllables and hence morphemes and words, and given a large enough vocabulary, it is not unlikely that some words will match even if there is no genetic relationship between two languages at all!

Ideally, what needs to be examined is the grammatical patterns of Japonic languages and their putative linguistic relatives, Altaic and Austronesian languages. One intriguing similarity between Japonic and Austronesian languages (especially those in the Oceanic branch of the family) is that both share the abovementioned penchant for syllables consisting of a consonant followed by a vowel (CV syllables). On the other hand, interesting similarities are observed between the grammatical patterns found in Japonic and Altaic languages. In particular, Japanese (and Korean) share with Altaic languages a number of typological features: the Object-Verb order, postpositions, relative clauses preceding the noun they modify, verb-final interrogative suffixes, agglutinative morphology, nominative-accusative case marking and lack of (in)definiteness marking. However, such general typological similarities may not be enough to establish a genetic link between these languages, but can be explained away by general typological patterns. Simply put, these properties -- especially, the Object-Verb order and postpositions -- correlate across languages from different families. Moreover, these two surface patterns may be reduced to the same underlying structure: a head (a verb, a postposition) following rather than preceding its complement.

To recap, the observed grammatical similarities between Japonic and Altaic languages may not be sufficient to postulate a historical, genetic link between the two language families. Thus, the jury is still out on whether Japanese traces its roots to an ancestral language from the Korean peninsula or from Taiwan/China. And this question is not without political consequences: the link to an ancestral group from the Korean peninsula has been invoked to justify the annexation of Korea and Manchuria before World War II, while after the war the link with the Jōmon culture (presumably, from Taiwan) has been emphasized. New linguistic studies, such as the one by Lee and Hasegawa, suggesting a closer (linguistic) connection to the newcomers from the Korean peninsula, may tip the scale again.