In three investigations of the common outbreeding grass Festuca ovina The geographic variation of alleles was studied. The aim was to determine the role this variability has on the probability that an allele is lost from a particular sample, (i) In the investigation of Skane, where 35 populations were studied for 12 independent alleles, it was found that the probability of loss as a function of an allele's overall frequency was increased compared to the null-hypothesis of its frequency being the same in all populations. This implies that genes in a particular sample are only partly informative about the genetic composition of the area as a whole. A method is described how to calculate this degree of representativity, R. In the investigation of Skåne, with an average of 33 analysed genes per sample, the estimated value for R was 0.68. This value can be taken as an upper estimate of R for other herbaceous plants, since the investigated system was chosen to be as homogeneous and have as high an R value as possible, (ii) From a study of 16 samples from the whole of Fennoscandia could be concluded that sampling populations with different post-glacial histories is an effective way of increasing the number of recorded alleles. (iii) A study of seeds from seven selected plants in a local population at Vomb in Skåne, showed that the allele frequencies vary significantly between the pollen clouds that fertilize different individual plants. Thus, in a plant as outbreeding as Festuca ovina, genetic variability is pronounced at the smallest and the largest geographic scales, but is weak at the intermediary scale.