This study is a treatment of the Narada dialect，spoken in a linguistically isolated village in central Japan. The previous treatments of this dialect are critically examined and the popularly accepted notion that ‘every lexical item in Narada has one accent' is rejected.
It is shown that this dialect has underlying accentual representations and phonological rules which are essentially identical with those of the neighboring dialects ( e.g.Koofu) and that the surface difference between these dialects can be attributed to differences in the Pitch Realization Rules. A set of such rules are proposed and illustrated.
Several accentual alternations in Narada are discussed，and it is shown that Narada and Koofu exhibit parallel alternations. The differences between the older generation
speakers and the younger speakers are discussed and a possible solution is suggested to account for the differences.