In order to study the genetic variation and relationships between seed yield and its components 36 populations of cocksfoot (Dactylis glomerata L.), were sown using complete block design with three replications in Hamadan agricultural research station. Iran in 2010. Analysis of variance showed significant differences among populations for dry matter (DM) yield, plant height, seed yield, panicle number, 1000-grain weight, heading date, flowering date, panicle length and harvest index. The estimates of broad sense heritability for the various traits were low to high (h2b=0.17 to 0.90). The results of phenotypic correlation showed a positive correlation between seed yield with DM yield, panicle number and harvest index. Using principal component analysis, the most important variables were identified. The first three independent components accounted for 46, 21 and 10% of total variation. The relative magnitude of eigenvectors from the first principal component axis, indicates that heading date, flowering date, plant height, panicle number, DM yield and panicle length were the important traits for classifying accessions into clusters. From the second principal component axis, seed yield and harvest index were important variables. 1000 grains weight had high eigenvectors in the third principal component axis. Based on Ward clustering method, 36 were grouped into three clusters. Populations in cluster 1 were known as late maturity group couple with lower yield. Populations in clusters 2 were defined as mid season group plus with higher seed production and cluster 3 averaged well above the overall mean for DM yield couple with early maturity. It was concluded that there was moderate to high heritability for most traits in the cocksfoot populations and selection for DM yield could result in increase in seed yield
