Estimating kernel characteristics in winter wheat (Triticum aestivum L.) using conventional methods is time consuming and destructive. An alternative non-destructive method for the estimation of these traits would facilitate early generation selection with large sample numbers or when time between harvest and planting is limited. The objective of this study was to develop and validate whole-grain calibrations using near-infrared reflectance (NIR) spectroscopy to estimate kernel weight, kernel diameter, and test weight. Two thousand thirty-eight whole grain samples were selected from early and late generation breeding nurseries and a recombinant inbred population grown at multiple locations throughout eastern Colorado. For each sample, NIR spectra (from 1100-1500 nm) and reference values were collected. Reference trait values were regressed against the first and second derivative transformations of spectra by modified partial least squares regression. Optimum equations were developed that had R² values of 0.80 for kernel weight and 0.76 for kernel diameter. Using these equations, selection efficiency was determined in early generation breeding material. Results reveal the possible utility and efficiency of using whole grain NIR spectroscopy in a winter wheat breeding program.