Was Japanese "made in Japan"? Since Japan was not the cradle of mankind, the first speakers of proto-Japanese must have come from elsewhere at any rate. Do they still have recognizable relatives there, at least linguistically? There are reasons to think so.
Recently I attended a lecture at my Alma Mater by Dr. Martine Robbeets updating her Ph.D. research on the linguistic roots of Japanese in a wider language family: Is Japanese Related to Korean, Tungusic, Mongolic and Turkic? (Harrassowitz, Wiesbaden 2005). She gave a convincing account of her thesis that Japanese is indeed cognate to a string of continental languages. At the risk of gross simplification, I will attempt to summarize what I retained of it.
The first thing to do in this kind of comparative exercise is to establish a list of non-trivial isoglosses between these languages, i.e. any linguistic similarities: syntactic, morphological, lexical. Trivial ones have to be discarded, starting with correspondences based on near-universals of language. Thus, the fact that words for "mother" have a characteristic [m] sound, as in Chinese mu or Dravidian amma, doesn't prove that the languages have a common origin, nor that one language borrowed the word from another, only that babies at their mother's breast make the same sound everywhere.
The same caution is needed against loan-words, e.g. Japanese biiru, garasu and karan sound like Dutch bier (beer), glas (glass) and kraan (watertap) not because the languages share an ancient common origin but because Japanese borrowed the Dutch words during the colonial age. Finally, we should guard against similarities based on pure coincidence, e.g. in the case of Japanese namae and its English counterpart name, words which have a distinct history in Japanese c.q. Indo-European. If we exclude such similarities from our survey, we can eliminate some languages that have been proposed as cognate to Japanese on this rather flimsy ground, such as Tamil and Sumerian.
Since a few decades, the group of Korean, Tongusic, Mongolic and Turkic is considered the most likely candidate for kinship with Japanese. The latter three have been grouped together as the "Altaic" language family. Strictly, it is called the "Altaic hypothesis", for the kinship between the three has not been firmly established. Given their overlapping territories, the languages may have mixed so much that distinct origins may have gotten obscured by the overlay of mutual exchanges. Those who acept the Altaic hypothesis tend to include Korean in the group as well. However, the name "Altaic" is now outdated and replaced with Trans-Eurasiatic (i.e. now stretching through the Eurasian continent from Korea to the Balkans). The Altai mountains and the surrounding steppes and deserts are an inhospitable region, unlikely to be a demographic epicentre of emigrations. Archaeology and common sense confirm that the homeland of "Altaic" must be in a more inhabitable area, viz. South Manchuria.
It is from there that these languages were spread westward (Turkic, Mongolic), northward (Turkic, Tungusic) and eastward (Korean, Japanese). Influenes from that region remain discernible in the archaeological remains of the immigrant Neolithic Jômon culture that replaced the older Yayoi culture. Physical anthropology also confirms that the Jômon people had more in common with those of continental Northeast Asia, while the Yayoi people resembled those of Southeast Asia. More precisely, Dr. Robbeets suggested that Japanese originated in South Korea, while modern Korean was originally the language of North Korea.
She lists a number of lexical correspondences, e.g. words for "hard" in Turkic, Mongolic, Korean and Japanese can be deduced from a common origin *kata. But the critical evidence is the list of correspondence between morphological markers. These are unlikely to be exchanged between languages; English has a few from French and Latin (e.g. the suffix -nce in comeuppance, or -able in likeable, and even these are not part of the intimate layer of morphology, such as verb conjugation), but it has been exposed for many centuries to an intense impact of these languages. Japanese, by contrast, has not in living memory been in contact with Turkic, Mongolian and Tungusic. So, when she finds the same suffix or infix present in all five language groups (or in one case four out of five), it is a strong pointer to deep kinship between the languages. This is all the more true when these suffixes are no longer recognizable as distinct entities, thus no longer available for borrowing, because they have been integrated into verbal stems as now operating in the language.
In trying to link Japanese to the four other groups, Dr. Robbeets off-hand demonstrates the kinship between the four others, and thus the Altaic hypothesis. If confirmed, this finding constitutes an important success for the historical-comparative method in language studies.
In my line of research, this is important because many in India deny the validity of this method, thinking it "guilty" of the Aryan Invasion Theory. Their reasoning is that a theory that leads to a wrong conclusion (viz. the Aryan Invasion hypothesis) thereby stands disproven. They, along with their pro-AIT opponents, are simply mistaken in thinking that the theory (based on the historical-comparative linguistics) of an Indo-European language family with a common origin necessarily implies that this common origin lay outside India. Positing a common origin and thus a common homeland implies in itself nothing about the location of that homeland. Likewise, in the case of the Altaic hypothesis its originators first vaguely associated it with the Altai region but have relocated the putative common homeland to South Manchuria, without therefore abandoning the hypothesis of a common origin and homeland, much less the linguistic method.