In plant populations alleles often deviate from a random distribution and reveal positive autocorrelation at short distances. In species with both clonal and sexual reproduction, such clustering may be because ramets of the same genet were sampled at nearby locations. Alternatively, clustering may be the result of limited gene flow through pollen or seeds (isolation-by-distance). Here, we modify a conventional spatial autocorrelation analysis using the join-count statistic in order to differentiate between these two causes of genetic structure. We examined the distribution of seven microsatellite loci representing 37 alleles in a 20 x 80 m plot of a perennial population of eelgrass Zostera marina L. In analysing join-counts between all like genotypes we found significant genetic autocorrelation among ramets at distances between 1 and 7 m (P < 0. 001). We then excluded joins between clonemates which were identified from the expected likelihood of their seven-locus genotypes. Without joins within genets, no autocorrelation was evident, indicating that most of the significant genetic clustering was caused by clonal spread. At distances up to 27 m, alleles were distributed at random, indicating a panmictic population at this spatial scale. These results illustrate the need for an a priori estimation of genet-ramet structure in clonally reproducing plants in order to avoid erroneous inferences about putative gene flow at various spatial scales.