It is difficult to assess the relative variability or stability of chasmogamous (CH) and cleistogamous (CL) reproduction in perennial herbs with mixed mating because long-term data in natural populations are unavailable. Here, the aim was to quantify and compare spatial (between-habitat) and temporal (among-year) variation in CH and CL reproduction over 5 years in two subpopulations of the native perennial grass Danthonia compressa. This species produces CH spikelets on terminal panicles in early summer, while axillary CL spikelets, including a basal cleistogene, mature into the autumn. Flowering tillers were collected from a sunny woodland edge and an adjacent shady interior habitat for 5 consecutive years (2017-21). Seed set, fecundity, seed mass and biomass allocation were recorded for the two floral types along with tiller vegetative mass. Bivariate line fitting was used for allometric analysis of CH and CL fecundity. Seed set, fecundity, mass per seed and allocation to seeds differed between floral types and showed significant variation between habitats and among years. Seed set and fecundity in CH panicles were greater than that of axillary CL panicles in most years. Tiller mass positively affected axillary CL seed production and mass of the basal cleistogene. Fecundity and allocation among years were more variable for CH compared to CL reproduction. High seed set and fecundity of CH spikelets suggest that pollination does not limit reproduction via chasmogamy. Late maturation of axillary CL spikelets provides additional fecundity, especially in larger plants along sunny woodland edges. The heavy cleistogene at the tiller base could be important to population persistence, analogous to the axillary bud bank of other perennial grasses that are not cleistogamous. The spatiotemporal stability of CL reproduction underscores the ecological significance of cleistogamy to reproductive fitness.