An understanding of how selfing occurs and whether selfing incurs costs relative to outcrossing is necessary for an evolutionary analysis of self‐fertilization. We used glasshouse and field experiments to address these questions in the perennial herb Bulbine bulbosa. Flowers did not autonomously self‐pollinate, and geitonogamy was unlikely because plants opened only one short‐lived flower per day on average. Instead, selfing occurred via facilitated autogamy. We manipulated selfing rate and showed that seed set was negatively related to levels of selfing because many selfed zygotes aborted. Selfed seedlings were smaller and fewer survived compared with crossed seedlings, resulting in high cumulative inbreeding depression (0.85). Naturally pollinated plants had high ovule fertilization similar to that of experimentally cross‐pollinated and self‐pollinated plants. However, naturally pollinated plants produced fewer fruits and seeds than did cross‐pollinated plants but similar numbers as self‐pollinated plan...