Grass pea (Lathyrus sativus L.) is relatively tolerant to several abiotic stresses such as salinity. Twenty-five grass pea genotypes provided by the Dryland Agricultural Research Institute, Maragheh, Iran, were exposed to four salinity levels (0, 40, 80, 120 mM). Based on the primary screen, 8 tolerant genotypes and 17 sensitive genotypes were selected. Principal components analysis for the sensitive and tolerant genotypes showed that the first four components explained approximately 71-80% of variance. In PCA, the coefficients associated with leaf showed that, in tolerant genotypes, leaf length and leaf number increased but, leaf width decreased. In other words, salinity stress makes the leaves grow larger and longer, but the width of the leaves decreases and angle of the leaf becomes tighter. So, these traits may be used as an index of salinity tolerance in grass pea selection programs. Cluster analysis based on Ward’s algorithm and Euclidean distance measure effectively separated the tolerant and sensitive genotypes. Cut off dendrogram located tolerant genotypes (except one genotype) in the same subgroup. This fact shows that cluster analysis efficiently divided and/or separated semi-tolerant tolerant genotypes. Shoot dry weight had significant negative correlation with pH, plumule length and plantlet length. Path analysis on tolerant genotypes, revealed the direct effects of leaf number and pod number with shoot dry weight. In stepwise regression, two traits including leaf number and pod number were entered into the model. The results showed that leaf number and pod number had the most direct and positive effect on the shoot dry weight, and the plant height had the highest indirect effect through leaf number on the shoot dry weight. About sensitive genotypes, the results showed that leaf number and fruit fresh weight had the most direct and positive effect on the shoot dry weight, and the plant height had the highest indirect effect through leaf number on the sh
