Mine is a two color experiment, where each of four chips is a biological replicate, and half of these are dye-swapped. The (biological) control is always in the first channel and the treatment is always in the other. My confusion must stem from the R/G mnenomic. Cy5 fluorescence is in the red and Cy3 in the green, but this is irrelevant. Limma's R/G refer to false color. My controls are always false colored red, and the experimentals always green, irrespective of the dye. Thus, all of the control values should go into one matrix, and all the experimental values into the other, and my design would be c(1,1,1,1). Half of the measurements in the R channel should be cy3 measurements and half should be cy5 measurements. That way, dye-biased measurements from the same spot, assuming no biological change, should average out to zero, after normalization. Does that sound right? Thanks, John [[alternative HTML version deleted]]