The spatial genetic substructuring of five local populations of Siebold's beech (Fagus crenata Blume) in the western Tanzawa Mountains were analyzed using allozyme polymorphisms for 13 loci (11 enzyme systems). The beech forests in Tanzawa have been rapidly deteriorating in recent years. The two main causes for this are heavy damage or death of mature trees caused by air pollution, and feeding pressure on seedlings and juveniles by sika deer. The distributions of genotypes showed localization of particular genetic diversities. Spatial autocorrelation analyses revealed slightly positive significant values within small distance classes and negative values in further distance classes, but those values more or less fluctuated in the distrograms. In addition, an analysis was conducted using a stepping-stone model to estimate the migration rate and to quantify the gene flow among neighboring patch populations to detect the population differentiation. Twelve patch populations were discriminated as localized assemblages of Siebold's beech individuals in one of the local populations according to the assemblage of individuals along the topography of an established site. The migration rate for each generation was estimated to be m = 0.09 between neighboring patch populations. The population migration rate (Nm = 2.7) was revealed to be larger than one migrant for every other generation, which indicates that the population differentiation is not significant in the present state. Although the model analysis demonstrated sufficient gene flow between neighboring patch populations, severe damage to local beech populations in Tanzawa may cause the extinction of one or more of the patch populations. Further fragmentation and isolation of local populations were also discussed in terms of loss and retreat of genetic diversity.