Floral trait correlations were compared between the two flower morphs of a distylous Ophiorrhiza napoensis population in a subtropical evergreen forest at the Defu Natural Animal Preserve, Guangxi, China. Common principal component analyses indicated that overall patterns in correlations among floral traits were morph specific in the study population. Strong positive correlations (r > 0.9) between anther height and corolla-tube length were found in both morphs. Stigma height correlated positively with corolla-tube length in the long-styled morph (r = 0.843), but not in the short-styled morph (r = −0.018). Flower-morph-specific correlation suggests that natural selection by pollinators has moulded trait covariance among floral traits. Because morph-specific correlations are expressed as the patterns of within-morph variation among multiple traits, putative genes responsible for the stigma-corolla tube correlation should not link to the supergene for sex-organ reciprocity between the morphs, but their expression is limited in the long-styled morph.

The purpose of this study was to describe morph-specific patterns of stigma–anther separation and style curling in the distylous Jasminum fruticans. We also examined whether variation in floral traits is correlated with variation in seed production. Stigma–anther separation is more variable in short-styled plants than in long-styled plants. In all populations studied, some plants bear flowers that lack any stigma–anther separation. These plants have the pollen characteristics (size and number) and compatibility relations of short-styled plants. Comparison with other distylous species illustrates that the variability of stigma–anther separation in short-styled plants is a novel finding for a distylous species. Long-styled plants have greater stigma–anther separation than short-styled plants, styles are often curled and protrude from the corolla, and anthers are placed well within the corolla tube. The frequency of long-styled plants with curled styles and the mean degree of style curling were significantly correlated with style length. Short-styled plants have larger corollas than long-styled plants and never have curled styles. Morph ratios are always 50[ratio ]50 in natural populations. Mean values of each floral trait in the two morphs were significantly correlated among populations. There were no consistent differences in fecundity of the two morphs nor any correlation between floral traits and seed set for each morph in natural populations. We discuss the potential causes and significance of the two morph-specific patterns we describe; reduced stigma–anther separation in short-styled plants and the presence of curled styles in long-styled plants.