Early prediction of grain yield in a breeding program would be is considered a very important factor for making selection. This study was undertaken to determine the potential of using spectral reflectance indices as an indirect selection tool for yield in wheat breeding programs. To achieve this objective, 4 different experiments, each consisting of 25 winter wheat genotypes, were conducted at the OSU Agronomy Research Farm. One trial was composed of 25 cultivars from the Southern and Central Great Plains, and the other three trials contained randomly F5 derived F6 sister lines from 3 different crosses. Six different spectral reflectance indices (three included a visible and a NIR region and the other three included the minor water absorption band) were calculated by using a portable FieldSpec spectrometer (Analytical Spectral Devices, Boulder, CO) that gives the reflectance spectra from 350 to1050nm wavebands with a 1.5 nm increment. Indices were calculated at three different growth stages (booting, heading, and grain filling) of the crop growing cycle. Reflectance measurement taken at the heading stage best predicted yield. Indices that included the minor water absorption bands were better correlated with yield than the commonly used vegetation indices based on both NIR and visible wavebands (RNDVI, GNDVI and SR). These trends suggest that the water status of the canopy is a very important determining factor for the productivity of a given genotype. When indices taken at different growth stages were considered together, the indices based on the water absorption band explained 45% - 70% of the yield variability in different trials. The relationship between the spectral reflectance indices and grain yield best fit a linear model, suggesting that the reflectance indices can be used in a breeding program for identifying the best yielding genotypes as well as the lowest performing genotypes.