In this study, fourteen markers were analyzed to assess the genetic diversity of four wild geographical populations of Liza haematocheila from the adjacent waters of Huludao, Qingdao, Lianyungang, and Zhoushan. The results showed that there was a certain degree of polymorphism in the four populations and the percentages of polymorphic loci were 92.86%, 92.87%, 100%, and 85.71% respectively. A total of 61 alleles were detected in the four populations, and the average number of alleles was between 3.786 and 4.000 for each population and the effective number of alleles was from 2.673 to 2.899. The average values of observed and expected heterozygosity ranged from 0.359 to 0.389 and from 0.503 to 0.561 respectively. The mean polymorphic information contents were from 0.465 to 0.513. These results suggested that there was a moderate genetic diversity among the four populations, however, the Kruskal-Wallis test (SPSS software) indicated no significant difference in the genetic diversity among the four populations (H=0.187, df=3, P=0.980). The majority of loci in each population were found to deviate from the Hardy-Weinberg equilibrium (P<0.05). Moreover the Huludao and Zhoushan populations were found to have two loci with heterozygote excess (Fis<0), while the other two populations had three loci with heterozygote excess. This was probably caused by recent bottleneck effect. The values of genetic differentiation coefficients and gene flow were 0.148 and 1.444 respectively, which suggested a medium level of genetic differentiation and a certain degree of gene flow between the four populations. The genetic similarity coefficients of the four populations ranged from 0.623 to 0.818 and the genetic distance was between 0.202 and 0.473 among the populations. The genetic distance was the nearest between Huludao and Qingdao, and was the farthest between Lianyungang and Zhoushan. These results could be related to the diffusion of the mullet larvae, the inshore coastal ecological environment, and the community structure. Cluster analysis with the UPGMA method showed that the four geographic populations could converge into one cluster according to their distribution from north to south.