Undeclared mustard residues in food products could trigger allergic reactions in mustard-allergic consumers. Our objective was to develop and validate a sandwich-type ELISA for the detection of mustard residues in foods. A mixture of yellow, brown, and oriental mustard seeds was used to immunize 3 rabbits and 1 sheep. Two mustard ELISAs were developed by utilizing the reciprocal combination of rabbit and sheep polyclonal antimustard sera as the capture and detector reagents. Binding was visualized by addition of rabbit antisheep or goat antirabbit IgG antibody labeled with alkaline phosphatase and subsequent addition of substrate. The optimized ELISAs have limits of quantification (LOQ) of 1 and 3 ppm (μg of ground, whole mustard seeds/mL) for the sheep capture and rabbit capture formats, respectively. Only rapeseed cross-reacted in the rabbit and sheep capture mustard ELISAs at a level equivalent to 12300 and 16900 ppm of mustard. The mean percent recovery for cooked frankfurters spiked with 0 to 1000 ppm mustard flour was 95.3% ± 10.7%. A limited retail survey of 29 foods revealed that, of 15 samples having mustard declared on the ingredient list, 2 baked bean products contained no detectable mustard, possibly owing to a decrease in extractability and detectability of mustard proteins after subjecting to thermal processing. For the remaining 14 samples without mustard declared on the label, 3 samples contained detectable mustard, presumably due to the labeling of mustard as "spice" or inadvertent cross-contamination. This sandwich-type ELISA can serve as a powerful tool for food manufacturers and regulatory agencies to detect and quantify mustard residues in processed foods.