This is a revised version of the lecture delivered at the 88th General Meeting of the Linguistic Society of Japan. The paper deals with theproblem of linguistic reconstruction from the standpoint of linguistictypology and language universals, taking up two examples in the fieldof phonology, namely, the stop system of Proto-Indo-European and thevowel system of Old Japanese.
As to the PIE stop system, the distributional and hierarchical relationsamong the phonemes in question cannot be adequately accounted for by their traditional interpretation as voiceless (Tenues), voiced (Mediae) and voiced aspirate (Mediae Aspiratae) respectively. Thus the writerconcludes, in agreement with the recent “Glottalic Theory”, that the traditional Mediae must be re-interpreted as glottalized or ejective stopsand the traditional Mediae Aspiratae as usual voiced stops.
As to the Old Japanese vowel system, the writer examines the reconstructionsand interpretations recently proposed by various scholars concernningthe so-called Mannyoo-Gana system and concludes that theproblem of i₁/i₂f and e₁/e₂ cannot be fully solved without consideringthe phenomenon as a entire diachronic process, which seems to have begunwith the vocalic distinction accompanied by the non-distinctive consonantaldifference (i. e. palatalized/nonpalatalized) of the syllablesin question and to have ended in their complete merger, while the graphemicdistinction of o₁/o₂ is not to be interpreted as phonemic oppositionconnected with vowel harmony, such as /o/ and /ö/, but rather asallophonic variation reflected on the foreign writing system. The vowelsystem of Pre-or Proto-Japanese is to be reconstructed, in the writer'sopinion, as the four vowel system with /a/ in medial position whichshifted later to /o/ as a result of the appearance of the new vowel /e/from *a-i and *i-a. And the vowel system of the 8th C., which seemsto have been in quite a transitional and fluid state, can be reconstructedas the five or six (if Ci₂ is interpreted as /Ci/) vowel system.