It is desirable for sorghum breeder to know the extent of relationship between yield and each of morphological and physiological traits affecting it which facilitate breeder in selecting plants of desirable traits, especially under drought stress conditions. Therefore, this work was conducted at two locations represent clay and sandy soils of the Experimental Farms of Faculty of Agriculture, Assiut University, Assiut, Egypt, during the two summer seasons 2014 and 2015 under well-watering (WW) and water stress (DS) conditions. Three statistical procedures, i.e. simple correlation coefficient, the path coefficient analysis and the stepwise regression analysis were performed to determine the functional relationships between grain yield (GY) and each of its attributes and some physiological traits of 43 grain sorghum genotypes under both treatments. Results revealed highly significant differences among genotypes for all studied traits under both conditions. Panicle weight (PWG) had the highest positive correlation with grain yield of sorghum genotypes under both treatments followed by threshing percentage (TH%) and seed index, reflecting these traits are the most contributed to yield. Path analysis showed that PWG and TH% had positive and direct effect on GY, while chlorophyll content (CC), excised leaf water loss (ELWL) and stay green were the most important physiological traits under DS conditions. In addition, panicle width and CC showed the highest positive indirect effects on GY. Stepwise regression exhibited that PWG and TH% had the strongest variation in GY under both water regimes. Further, all physiological traits except excised leaf water loss (under WW) showed positive correlation coefficient with GY under both treatments. Also, relative water content was the most important physiological trait followed by flag leaf area under WW conditions, while CC was the most important physiological trait under DS followed by ELWL that contributed high amount of the total variation of grain yield.