Castanopsis eyrei is one of the main trees found in the subtropical evergreen broad-leaved forests of China's Zhejiang Province. The non-random genetic variation that describes the spatial genetic structure in this species reflects the ecological and evolutionary processes of the plant population. Spatial autocorrelation analysis is an effective method to study the spatial structure of genetic variation. The autocorrelation coefficient provides a measure of whether the location falls within a specific distance class for genetic similarity among individuals. Here we analyzed the fine-scale spatial genetic structure of the individual plants in C. eyrei populations and studied the individuals classified by age using the spatial autocorrelation method to quantify spatial patterns of genetic variation within the populations and to explore potential mechanisms that determine genetic variation in the plant populations. The spatial autocorrelation coefficient (r) at 10 distance classes was determined on the basis of both geographical distance and with a genetic distance matrix which was derived from ISSR binary data. All analyses were conducted using GenAlEx software. The results showed that all the individuals of C. eyrei exhibited significantly positive spatial genetic structure at distance less than 10m (the X-intercept was 9.945). This indicated that the same genotype of C. eyrei when grouped together had a small mean length within small genetic patches. Limited seed dispersal was found to be the main factor that leads to the finding that genetic variation within populations adopted a certain spatial structure. For species whose seeds are spread by gravity, the spatial correlation between individuals would decrease as the distance increased. In C. eyrei populations on the rough and uneven slopes that form the Shiwo terrain in the mountain, the secondary spread of the seeds is limited to a small area. In addition, forest litter covering the ground limits the secondary spread of the mature seeds to around the mother tree. Therefore, within these smaller geographical distances the genotypes were similar. Thus, significant spatial genetic structures are formed at short distances and weak spatial genetic structures are formed at long distances. The spatial genetic structure might be the result of seed gravity dispersal and extensive pollen dispersal. The individuals of C. eyrei in the three age classes, age class I, age class II and age class III, showed significantly positive spatial genetic structures at distance less than 10 m. The spatial autocorrelation coefficients at 5 m were 0.068, 0.054 and 0.070 for the age classes I, II and III, respectively, and for the three age classes the X-intercepts were at 11.820, 9.746 and 9.792, respectively. The individuals in age class IV showed no significant spatial genetic structure in any of the spatial distance classes. C. eyrei is a long-lived perennial plant and the evergreen broadleaf forest that C. eyrei was in was well preserved with little human interference. The loss of spatial genetic structure for C. eyrei in age class IV might be due to self-thinning. The age class IV population had no significant spatial structure, indicating that the self-thinning process was genotype non-dependent; cohort competition between individuals may lead to self-thinning. To further understand the survival and maintenance mechanism of C. eyrei populations, an in-depth study of the breeding system and the seed and pollen dispersal patterns is required.