To examine whether the Janzen–Connell hypothesis is valid in temperate forests, we investigated the density, growth, mortality, and agents of mortality of seedlings and the density, size, and age of saplings of Prunus grayana Maxim. at three distances (0–3, 6–10, and 16–26 m) from conspecific adults in a temperate forest in Japan. An inoculation experiment was also conducted to test the host range of a leaf pathogen. The probability of mortality was highest at 0–3 m during the first 2 years of growth. Mortality mainly resulted from distance-dependent attack by two types of pathogen that caused damping-off epidemics and spot symptoms on leaves. The leaf pathogen was identified as Phaeoisariopsis pruni-grayanae Sawada, which infected many more seedlings of Prunus grayana than of the two other tree species tested in an inoculation experiment. The vertical and diameter growth was lowest at 0–3 m and highest at 16–26 m in both seedlings and saplings. As a result, the greatest number of large and older saplings was observed at 16–26 m. Our results demonstrate that the Janzen–Connell mechanism operates in a beech-dominated forest in the temperate region of Japan.

To evaluate the extent to which landslides affect community dynamics and consequent species diversity in a beech‐dominated forest, differences in the composition and size structure of tree species were compared between landslide and adjacent stable (control) stands. Demography and changes in size were compared between the two stands over a 5‐year period about 60 years after a landslide. In the control stand, replacement occurred even amongst late‐successional species, with beech (Fagus crenata)—the most dominant species—increasing in relative abundance. In the landslide stand, very few large individuals of late‐successional species occurred, whereas large individuals of early‐successional species occurred only in the landslide stand. The traits indicate that the landslide strongly facilitated species diversity, not only by reducing the dominance of late‐successional species, but also by promoting recruitment of early‐successional species. However, new recruitment of early‐successional species was inhibited in the landslide stand, although we observed succeeding regeneration and subsequent population growth of late‐successional species there. As a result, the relative dominance of late‐successional species increased with succession after the landslide, thus decreasing future species diversity. In beech‐dominant forest landscapes in Japan that include communities with different developmental stages, the mosaic of serial stages may facilitate species diversity after a landslide.