Experimental work was conducted chiefly in an igloo, half way up Mt. Taisetsu in Hokkaido, at an altitude of 1050 m. Snow crystals were received on the collodion film of the holder of an electron microscope, and left in a desiccator kept between −4 and −8C. The crystals sublimed, and the supposed nuclei remained on the collodion film. These specimens were brought under electron-microscope investigation. One solid nucleus (center nucleus) was always observed in the central portion of a snow crystal. The center nuclei were between 0.5 and 8µ in largest extent. In the other parts of snow crystals, numerous smaller nuclei (condensation nuclei) were observed. These were nearly the same in size as condensation nuclei in the free atmosphere. 43 successful photographs of center nuclei were obtained. Most of them are considered to be soil particles, some being a lump of carbon particles, a micro-organism, or a hygroscopic particle of some chemical compound. 60 photographs of condensation nuclei were taken, and a frequency curve of size was made from 1200 data. Condensation nuclei are found to be of two kinds, the larger ones most frequently having a diameter of about 0.15µ, the smaller ones about 0.05µ. The mechanism of snow-crystal growth is discussed on the basis of these data.