Corn (Zea mays L.) grown for different end-uses is becoming a common strategy to increase farmers' revenue. Different end-uses (extractable starch, food-grade, etc) require particular grain quality characteristics, with the concentrations of starch, oil or protein being of particular importance. Environmental and economic pressures on production agriculture have increased the interest in variable rate Nitrogen (N) fertilization as a means to improve nitrogen use efficiency and grain composition, particularly protein. Spatial variability in the response to N is a prerequisite for variable rate N fertilization. If the levels of protein, starch or oil in the grain respond to N homogeneously within fields then variable rate fertilization is not a useful technology to improve this grain quality attribute. Because information on the spatial variability of grain composition responses to N fertilization is limited, it is difficult to asses the potential value of variable rate nitrogen fertilization. Thus, the main objective of this study was to analyze the spatial variability of corn grain yield, protein, oil, and starch responses to N fertilization within commercial fields. On-farm experiments were conducted during the 2003, 2004, and 2005 growing seasons in two growing areas in Illinois, USA. Five (2003) and six (2004 and 2005) N fertilizer rates ranging from 0 to 300 kg N ha^-1 were applied with commercial applicators in five to ten areas within these fields. At physiological maturity corn ears were sampled at four georeferenced points within each plot and yield, grain protein, oil, and starch concentrations were determined. The response of these variables to N fertilizer was analyzed by polynomial regression adjusted to account for the spatial autocorrelation of the errors. Yield response to N varied both between fields and within fields. Within almost all fields, corn yield did not respond to N in some areas but showed a large response in others. The N rate required to maximize grain yield and profit varied from 0 to 300 kg N ha^-1. The grain yield produced without N was more variable (for example 6.3 to 13.3 Mg ha^-1 in one field; 27% CV) than yield achieved with the optimal N rate for each area (12 to 16 Mg ha^-1; 12%CV). Grain protein concentration responded positively to N fertilization and this response also varied within fields, but less than grain yield. Some non-responsive areas in grain yield did not respond to N in grain protein, but others did. Grain starch concentration tended to decrease with N fertilization in most fields. On average across fields, grain starch concentration decreased 1.5 percent units with 300 kg N ha^-1 compared the unfertilized treatment. The spatial variability in starch response within fields was lower than for protein, and in most fields starch concentration decreased similarly, although in a few cases it was unaffected by N fertilization. Grain oil concentration was the grain component least affected by N fertilization, both at a field level and within fields. In some fields it did not respond to nitrogen fertilization, whereas in others it slightly decreased. The spatial variability of N response decreased in the following order: grain yield, protein, starch, and oil. The identification of areas where grain protein concentration does not respond to N, and of areas that show a large response, would allow for maximize grain protein in the harvested grain while minimizing the amount of N fertilizer applied. Similarly, detecting areas where starch concentration is minimally affected by N fertilization would allow for N to be applied based on yield, while still maintaining the starch concentration. The effect of N fertilization on oil concentration seems to be small and variable rate fertilization does not appear to be a promising tool to manage oil levels. These data show that variable rate N fertilization can improve grain composition and quality, particularly for traits associated with protein concentration, and become a management tool for producers of this type of specialty grain.