Abstract: Production of multiple flowers in inflorescences allows the reproductive phenotypes of individual plants to include systematic among-flower variation, which could be adaptive. Systematic trait variation within inflorescences could arise from resource competition among flowers, or be a developmentally determined feature of flower position, regardless of resource dynamics. The latter, architectural effect typically manifests as continuous floral variation within inflorescences. For architectural effects to be adaptive, floral trait variation among individuals must covary with reproductive performance and be heritable. However, heritability and phenotypic selection on gradients of variation cannot be estimated readily with traditional statistical approaches. Instead, we advocate and illustrate the application of two functional data analysis techniques with observations of Delphinium glaucum (Ranunculaceae). To demonstrate the parameters-as-data approach we quantify heritability of variation in anthesis rate, as represented by the regression coefficient relating daily anthesis rate to inflorescence age. SNP-based estimates detected significant heritability (h² = 0.245) for declining anthesis rate within inflorescences. Functional regression was used to assess phenotypic selection on anthesis rate and a floral trait (lower sepal length). The approach used spline curves that characterize within-inflorescence variation as functional predictors of a plant's fruit set. Selection on anthesis rate varied with inflorescence age and the duration of an individual's anthesis period. Lower sepal length experienced positive selection for basal and distal flowers, but negative selection for central flowers. These results illustrate the utility and power of functional-data analyses for studying architectural effects and specifically demonstrate that these effects are subject to natural selection and hence adaptive.

Key words: architectural effects, functional data analysis, heritability, inflorescence display, selection