The island syndrome for pollination systems in the Ogasawara Islands was evident in a high percentage of dioecious species, the subdued colour of the native flora and solitary flower visitors on satellite islands. The shape and colour adaptations of several flowers suggested native pollination niches for long-proboscis moths and carpenter bees. However, the domination and expansion of introduced honeybees have the potential for disruption of the native pollination network in the two main, and several satellite, islands of the Ogasawara Islands.

We analysed the regeneration process of Magnolia obovata using polymorphic microsatellite markers. Eighty-three adult trees standing in a watershed covering an area of 69 ha, and saplings collected from a smaller research plot (6 ha) located at the centre of the watershed were genotyped using microsatellite markers. Among 91 saplings analysed, 24 (26%) had both parents, 31 (34%) had one parent and 36 (40%) had no parent within the watershed. The proportion of genes in saplings inherited from the adults within the watershed was 43%, and therefore 57% were from outside the site, indicating active gene exchange across the watershed area. Average distance between parents and saplings (264.6 +/- 135.3 (SD) m) was significantly smaller than that of pairs randomly chosen between adults and saplings (436.7 +/- 203.0 (SD) m). The distance of pollen movement inferred from the distance between the two parents of each sapling ranged from 3.2 m to 540 m with an average of 131.1 m +/- 121.1 m (SD). Because 34% ( = 31/91) of saplings had only one parent within the watershed, the estimate of average pollen movement must be smaller than the actual one. Long-distance seed dispersal by birds, inbreeding depression and limitation in acceptance of pollen because of the difference of phenology in each individual flower were considered to be the probable causes of large gene exchange across the watershed.

Flowering activity and sex expression of Bischofia javanica Blume were investigated for 3 years. B. javanica is an invasive dioecious tree of subtropical forests on the Bonin Islands in the western Pacific of Japan. The sex ratio showed a significant male bias (1.25-2.33). Smaller trees were significantly male biased, whereas larger trees showed no significant difference in sex expression, suggesting that males tend to be more precocious in sexual reproduction. We found evidence for sex changes in B. javanica; these have not been reported previously. Most of the 1,653 census trees remained non-flowering (58.1 %); 3.7 % of them showed sex changes, and the percentage of trees repeatedly flowering as males and females was 10.5 and 3.4 %, respectively. Sex changes were observed in both directions but a larger percentage of male trees became female. Flowering frequency and sex expression were significantly related to tree size (i.e. diameter at breast height). Over the 3 years, trees that were consistent females were the largest; inconsistent trees (switching sex between years) were intermediate in size, whereas consistent males were the smallest. There were no significant differences in relative growth rate (RGR) among trees of different sex or flowering frequencies. These results suggest that the maintenance of female reproduction is not related to changes in RGR of diameter but to flowering frequency or the reversal to the male form, dependent upon the internal resource status of individual trees.

I investigated the effects of display size and flowering phenology on fruit set in Aucuba japonica, an understory dioecious shrub pollinated by opportunistic insects. Natural variations in display size, flowering phenology, and fruit set were monitored in 1997. A hand-pollination experiment was also conducted to check whether pollen limitation was a factor in fruit set in the field. Increases in floral display size did not affect fruit set; the proportion of flowers that set fruit was almost constant irrespective of the total number of flowers per inflorescence, the total number of inflorescences per plant, and the total number of flowers per plant. The hand-pollination experiment showed that fruit set was not pollen limited despite the low mating probability that resulted from the combination of dioecism and the species' dependence on opportunistic pollinators. This was due, in part, to the fact that female flowers did not have a predetermined period of receptivity, but instead remained receptive until they received pollen. In contrast, flowering phenology did affect fruit set. Fruit set was most abundant when male and female flowering was most abundant. This suggests there was some degree of pollen limitation during the part of the flowering season when male flowers were scarce.

The alien predatory lizard, Anolis carolinensis, has reduced the insect fauna on the two main islands of the Ogasawara archipelago in Japan. As a result of this disturbance, introduced honeybees are now the dominant visitors to flowers instead of endemic bees on these islands. On the other hand, satellite islands not invaded by alien anoles have retained the native flower visitors. The effects of pollinator change on plant reproduction were surveyed on these contrasting island groups. The total visitation rates and the number of interacting visitor groups on main islands were 63% and 30% lower than that on satellite islands, respectively. On the main islands, the honeybees preferred to visit alien flowers, whereas the dominant endemic bees on satellite islands tended to visit native flowers more frequently than alien flowers. These results suggest that alien anoles destroy the endemic pollination system and caused shift to alien mutualism. On the main islands, the natural fruit set of alien plants was significantly higher than that of native plants. In addition, the natural fruit set was positively correlated with the visitation rate of honeybees. Pollen limitation was observed in 53.3% of endemic species but only 16.7% of alien species. These data suggest that reproduction of alien plants was facilitated by the floral preference of introduced honeybees.