AbstractBackground and AimsGenetically controlled self-incompatibility (SI) mechanisms constrain selfing and thus have contributed to the evolutionary diversity of flowering plants. In homomorphic gametophytic SI (GSI) and homomorphic sporophytic SI (SSI), genetic control is usually by the single multi-allelic locus S. Both GSI and SSI prevent self pollen tubes reaching the ovary and so are pre-zygotic in action. In contrast, in taxa with late-acting self-incompatibility (LSI), rejection is often post-zygotic, since self pollen tubes grow to the ovary, where fertilization may occur prior to floral abscission. Alternatively, lack of self fruit set could be due to early-acting inbreeding depression (EID). The aim of our study was to investigate mechanisms underlying the lack of selfed fruit set in Handroanthus heptaphyllus in order to assess the likelihood of LSI versus EID.MethodsWe employed four full-sib diallels to study the genetic control of LSI in H. heptaphyllus using a precociously flowering variant. We also used fluorescence microscopy to study the incidence of ovule penetration by pollen tubes in pistils that abscised following pollination or initiated fruits.Key ResultsAll diallels showed reciprocally cross-incompatible full sibs (RCIs), reciprocally cross-compatible full sibs (RCCs) and non-reciprocally compatible full sibs (NRCs) in almost equal proportions. There was no significant difference between the incidences of ovule penetrations in abscised pistils following self- and cross-incompatible pollinations, but those in successful cross-pollinations were around 2-fold greater.ConclusionsA genetic model postulating a single S locus with four S alleles, one of which, in the maternal parent, is dominant to the other three, will produce RCI, RCC and NRC full sib situations each at 33 %, consistent with our diallel results. We favour this simple genetic control over an EID explanation since none of our pollinations, successful or unsuccessful, resulted in partial embryo development, as would be expected under a whole-genome EID effect.