Fifty diverse genotypes of bread wheat were evaluated in a randomized block design with three replication for the study of selection indices during Rabi, 2013. Thirtyone selection indices involving grain yield per plant (X1) and four yield components viz., biological yield per plant (X2), number of grains per main spike (X3), 1000-grain weight (X4) and grain weight per main spike (X5) were constructed using the discriminant function analysis. Discriminant function analysis indicated that selection efficiency of the function was improved by increasing the number of characters in the index. Among the single character indices, number of grains per main spike exhibited higher genetic advance and relative efficiency over straight selection for grain yield per plant. The index based on five characters viz., grain yield per plant, biological yield per plant, number of grains per main spike, 1000-grain weight and grain weight per main spike recorded the highest genetic advance as well as relative efficiency and selection efficiency. These characters could be advantageously exploited in the bread wheat breeding programmes.