This paper presents the results of a 5—yr field study on the determinants of individual variation in maternal fecundity (seed production) in the narrowly endemic violet Viola cazorlensis (Violaceae), at a southeastern Spanish locality. Flowers of this species are characterized by a very long, thin spur and broad morphological variability, and are pollinated by a single species of day—flying hawk moth (Macroglossum stellatarum; Lepidoptera, Sphingidae). The primary aim of this investigation was to answer the question, What are the relative importances, as explanations of individual differences in fecundity, of variability in floral traits and of other fecundity determinants that are of an extrinsic nature, such as microhabitat type and interactions with herbivores? The floral morphology of individual V. cazorlensis plants was characterized by means of both "conventional," linear measurements of the size of flower parts (petals, spur, peduncle), and shape analysis of corolla outline (using thin—plate splines relative warps analysis). Spatial (among substrate types) and temporal (among years) patterns of variation in flower, fruit, and seed production by V. cazorlensis plants are described, with particular emphasis on the comparative effects of floral morphology, herbivory (by mammalian ungulates and two species of lepidopteran larvae), and substrate type (rock cliffs, bare rocks at ground level, and sandy soils), on cumulative seed production at the individual plant level. Cumulative seed production of individual V. cazorlensis plants depended significantly on average floral morphology (both size and shape components), thus revealing the existence of phenotypic selection on the floral morphology of this species at the study population. Among all the floral traits examined, spur length was the only one for which no significant relationship with fecundity was found. Type of substrate largely determined differences between V. cazorlensis plants in the impact of herbivory (plants growing on the soil exhibited the greatest reproductive losses to herbivores), and it also influenced plant size and flower production per reproductive episode. Plant size, in turn, influenced the supra—annual frequency of flowering and the number of flowers produced in each reproductive event. Flower production and herbivory levels significantly influenced (positively and negatively, respectively) fruit number, which was the major direct determinant of seed production. Path analysis revealed that the main determinants of individual variation in cumulative seed production over the study period were, in decreasing order of importance (absolute value of "effect coefficient" in parentheses), cumulative fruit production (0.946), mean flower production per reproductive event (0.868), plant size (0.441), herbivory by ungulates (—0.221), and average score on the first relative warp (0.107), a descriptor of flower shape. After accounting for the effects of substrate type, herbivory, plant size, and flower and fruit production, individual variation in floral morphology (aspects of size and shape) explained a negligible proportion (2.1%) of total individual variation in cumulative fruit production. Phenotypic selection on the floral morphology of V. cazorlensis at the study population, although statistically significant, was therefore almost inconsequential as a source of individual variation in maternal fitness, its effects being heavily "dilute" by the overwhelming influence of other factors. As exemplified by this study, selection on the floral phenotype may often become largely irrelevant in evolutionary terms because other ecological factors are far more important determinants of fitness differences among plants. A realistic assessment of the potential relevance of selection on plant reproductive traits thus requires a quantitative evaluation, in its natural scenario, of the predictable consequences of such selection.