Field and ground layer vegetation in a mature Norway spruce (Picea abies L.) forest of Myrtillus-type in Kuorevesi, southern Finland, was sampled on 31 plots of sizes 0.5, 1 and 2 m(2) to find optimal balance among sampling effort, number of observed species, and accuracy of mean species-cover estimates. Rarefaction and bootstrapping were used in the data analysis. According to the results, the addition of species is slow yet continuous after the first few plots. The expected number of species in samples taken with the different plot sizes decreases in the following order: 2 > 1 > 0.5 m(2). Results implied that plant species have different optimum plot sizes with respect to the accuracy of the mean cover estimates. However, representative registering of species present needs a larger sample and plot size than a relatively accurate estimation of the mean cover of the individual species. Consequently, representative registering of species sets the minimum limits for the number and size of the sample plots.