Résumé
Self-fertilization is widely believed to be an ''evolu-tionary dead end'' [1, 2], increasing the risk of extinction [3] and the accumulation of deleterious mutations in genomes [4]. Strikingly, while the failure to adapt has always been central to the dead-end hypothesis [1, 2], there are no quantitative genetic selection experiments comparing the response to positive selection in selfing versus outcrossing populations. Here we studied the response to selection on a morphological trait in laboratory populations of a hermaphroditic, self-fertile snail under either selfing or outcrossing. We applied both treatments to two types of populations: some having undergone frequent selfing and purged a substantial fraction of their mutation load in their recent history [5], and others continuously maintained under outcrossing. Populations with a history of outcrossing respond faster to selection than those that have experienced selfing. In addition, when self-fertilization occurs during selection, the response is initially fast but then rapidly slows, while out-crossing populations maintain their response throughout the experiment. This occurs irrespective of past selfing history, suggesting that high levels of inbreeding depression, contrary to expectation [6], do not set strong limits to the response to selection under inbreeding, at least at the timescale of a few generations. More surprisingly, pheno-typic variance is consistently higher under selfing, although it quickly becomes less responsive to selection. This implies an increase in non-heritable variance, hence a breakdown of developmental canalization [7] under selfing. Our findings provide the first empirical support of the short-term positive and long-term negative effects of selfing on adaptive potential.