Oil yield and agronomic properties are the most important targets for canola breeders ( Brassica napus L.) in northern Europe. In order to enhance Pioneer's canola oil yield breeding efforts, oil content was measured with near infrared (NIR) spectroscopy directly on a harvester (NIR spectroscopy on-combine) and, in addition, moisture, protein and glucosinolate content were determined. NIR spectroscopy, coupled with rapid harvesting, can significantly improve the quality and speed of oil yield determinations so they occur within a very narrow harvest time window. Moisture, oil, protein and glucosinolate calibrations were developed with 449 samples from the 2004 to 2005 harvests, comprising spectra from four diode array NIR spectrometers mounted on-combine. Applying these calibrations to an independent dataset from the 2006 harvest resulted in standard errors of prediction ( SEP) and coefficients of determination ( r2) of 0.41% and 0.93 for moisture, 0.7% and 0.84 for oil, 0.61% and 0.81 for protein, 4.0% and 0.22 for glucosinolates, respectively. Combining calibrations generated from the four instruments gave optimal predictions. Omitting data from any instrument decreased accuracy and precision, although dropping each instrument had a different effect on the measured values of the constituents. Each instrument produced very similar moisture and constituent predictions with a common sample set as indicated by high r2 and, thus, very similar ranking properties. Analysis of variance with the on-combine determinations led to lower residual variance for oil and similar variance for protein compared to those obtained with classical methods of sampling and laboratory NIR analyses. In summary, the results demonstrate that NIR spectroscopy on-combine is very promising to enhance breeding canola for higher oil yield.