The medaka (Oryzias latipes), a tiny fish, has been an excellent experimental model for molecular and developmental genetics, and is expected for being a "vertebrate" model animal for population genetics, because 1) there is abundant within-species variation, and 2) the whole genome sequence has been determined for one of the inbred strains. In spite of its potential usefulness, there is no comprehensive study on quantifying between- and within-population genetic diversity of wild medaka. To investigate population structure, we examine nucleotide sequences of the non-coding D-loop region and the cytochrome b gene of the mitochondrial (mt) genome for medaka individuals collected with distinct two sampling-methods. Using deme-based sampling, out of 373 total individuals from three local (two wild and one biotope) populations only 16 distinct sequence types of mt D-loop are found. However, we find 26 D-loop types in 35 individuals collected with grid-based sampling from various geographical regions in East Asia. We carry out statistical tests to evaluate the distribution pattern of nucleotide frequencies among segregating sites under the standard neutral model, and we show that the deme-sampled populations might have experienced population size reduction and/or sub-structuring caused by natural or artificial migration. The reduction of genetic variation has been driven more markedly in the populations from the biotope, suggesting that human activities can pose an impact on the demographic history of medaka. Nevertheless, our results suggest that the grid-based sampling gives more abundant variations than the deme-based sampling, while the deme-based sampling gives more information about the local-wild characteristics of allele frequency spectrum than grid-based sampling. This could be a basis of strategy to use medaka as a vertebrate model animal for comparative population genomics.