Bidirectional selection on threshold size for flowering Cynoglossum officinale (hound's-tongue)

Previous studies have demonstrated that many facultative biennials show variation in threshold size for flowering. In order to quantify the genetic variation for this character, we performed a two-way artificial selection experiment on threshold size in Cynoglossum officinale. The parental generation, established from seed from a natural population, showed large variation in threshold sizes (2.6-13.4 g). After one generation of selection for low threshold sizes all plants in the F1 flowered at sizes below 3.2 g. In the high selection line none of the F1 plants under 3.6 g flowered, and there were very large nonflowering plants (up to 17.2 g). By interpreting the logistic regression of flowering probability on plant size as a cumulative frequency distribution of threshold sizes, we could derive the frequency distributions of threshold sizes in each generation. These were used to estimate the narrow-sense heritabilities by the standard procedure for artificial selection on a quantitative character. This gave 'heritabilities' of 0.35 and 0.32 for the high and low selection lines, respectively. A previously constructed optimization model, applied to the population in our study area, predicted strong selection against low threshold sizes, but a fairly equal fitness for threshold sizes around and above the optimum. This flat fitness profile may explain why the natural population of Cynoglossum officinale harbours extensive genetic variation for this character.