We used the mark-recapture method to gather fundamental information on population dynamics of the Japanese water shrew (Chimarrogale platycephala) in a mountain stream from April through December 2004. Trapping was conducted over 2,700 nights, during which 58 captures of 24 individuals occurred. The numbers of individuals were higher in the upper stream reach where few modifications were made, and lower downstream where dams of several meters in height and concrete-lined wall were located. Of 23 individuals (excepting one drowned individual), eight were not recaptured, six were recaptured within one month, and nine were recaptured within two to seven months after initial capture. These results suggest that the population size of the Japanese water shrew has been negatively affected by river modifications, and there are specific tendencies in their movement pattern.