Abstract: At present study, the reasons of “horse-shoe effect” in correspondence analysis for analyzing human population genetic structure was explained. Based on the structure of gene frequency matrix, we displaye the different patterns of Scallergram of correspondent analysis from different types of loci (HLA-A locus, and STR- CSF1PO locus in Chinese Han populations). The results indicate that different types of loci showed different patterns of Scallergram of correspondent analysis. When some alleles have very low frequency in the gene frequency matrix, there would be “horse-shoe effect” in the Scallergram of correspondent analysis. The reason is that the c2 distance measurement in correspondent analysis usually overrates the effect of the genes with low frequencies. To carry out the correspondent analysis of human population genetic structure, when the Scallergram presents “horse-shoe effect”, one should examine the structure of gene frequency matrix, and confirm whether the “horse-shoe effect” shows the real pattern of population genetic structure. Only in this way, one can explain the “horse-shoe effect” correctly.