Abstract: To investigate the shell thickness phenotypic variation and its application in walnut breeding, nine biological traits including vertical diameter, transverse diameter, lateral diameter, fruit mass, kernel mass, kernel rate, fatty acid content, protein content and shell thickness were collected, and their correlations were analyzed. Cluster analysis and principal component analysis (PCA) were adopted to study the population structure effects on the phenotypic variation of shell thickness in walnut. Among 66 walnut varieties, the variable coefficient of the shell thickness ranged from 0.60 to 1.50, and enjoyed the maximum variable coefficient of 0.21 among all traits. Correlation analysis indicated that the shell thickness was negatively correlated to kernel rate with r=-0.61 (P=0.01). The population was divided into four pop-populations (Q1, Q2, Q3 and Q4) with 42, 18, 5 and 1 in Q1, Q2, Q3 and Q4, respectively. The principal component analysis of 66 accessions based on the nine traits indicated that the first three principal components (PC1, PC2 and PC3) could explain 77% of the total phenotypic variation. The results can provide a theoretical guidance for paper shell cultivar breeding.