Scylla paramamosain (Decapoda: Portunidae) is a swimming crab that is widespread in the coast of Southeastern China and commonly found in estuarine and mangrove waters. Although it had been reported that no significant morphological and isozyme differences were found among populations of S. paramamosain in China, it is still necessary to study the genetic structure of S. paramamosain by more sensitive and stable molecular markers. Genetic diversity and structure of 10 populations of S. paramamosain were investigated and anglicized at the mtDNA lebel in this study. A total of 130 crabs from ten representative localities along the coast of Southeastern China were collected. A 522 bp length fragment of the mitochondrial DNA cytochrome oxidase subunit I (mtDNA COI) gene was amplified and sequenced. Intraspecific variation of mtDNA COI gene was investigated in 130 individuals. We examined 522 base pairs (bp) and identified 21 different haplotypes. Each population sample is characterised by a single most frequent haplotype (Hap 2) with the highest frequency of 56.15％, shared among all ten populations, and a small number of rare ones, typically present in only one or two individuals and representative of several specific populations. The average haplotype (h) and nucleotid diversity (π) of ten populations is 0.6738 and 0.1987%, respectively. Analysis of molecular variance (AMOVA) and F_ST statistics analysis of mtDNA haplotype frequencies and sequence divergence data revealed a significant genetic differentiation (F_ST＝0.05, P<0.005) among populations. In pairwise population comparisons, a highly significant genetic differentiation exists only between population BH and other 5 populations based on F_ST values from haplotype frequency and distance method, while population SY differs pollution HSL and TS significantly. Using a Mantel test, we found no significant correlation between pairwise population F_ST values (estimated by haplotypic frequencies) and geographic distances, suggesting that a simple model of isolation by distance is not appropriate for our data. Tests of neutral evolution (Tajima’s D and Fu’s Fs) revealed the D statistic was negative with a significant deviation from mutation-drift equilibrium, and result of Fu’s Fs-test, which was devised specifically to detect population expansion and is more sensitive to the presence of singletons in a sample, showed a highly significant value.