Abstract: A total 5963 of germplasm collection of upland cotton accessions were sampled for constructing a core collection. The whole collection was firstly divided into 11 groups based on the main the traits and genotypes. The resultant core collection including 281 accessions was constructed with 21 traits in each subset by unweighted pair-group method with arithmetic means (UPGMA), which accounted for 4.71% of total collection. Furthermore the representativeness between core collection and total collections was assessed by the t test, F test of variance, coefficient of variation, the diversity index t test, average, range, phenotypic variance and variation coefficient, mean difference percentage, variance difference percentage, coincidence rate of range, changeable rate of coefficient of variation and principal component analysis. The results demonstrated that the selected core collection could represent the genetic diversity of initial collection.