Gene expression profiling by DNA microarrays has found wide application in many fields of biomedical research. The protocols for this technique are not yet standardized, and for each given step in microarray analysis a number of different protocols are in use. As a consequence, results obtained in different laboratories can be difficult to compare. Of particular importance in this respect are the methods for the preparation of fluorescent cDNA probes that should quantitatively reflect the abundance of different mRNAs in the two samples to be compared. Here we systematically evaluate and compare five different published and/or commercial principles for the synthesis offluorescently labeled probes for microarray analysis (direct labeling, 77 RNA polymerase amplification, aminoallyl labeling, hapten-antibody enzymatic labeling, and 3-D multi-labeled structures). We show that individual labeling methods can significantly influence the expression pattern obtained in a microarray experiment and discuss the respective benefits and limitations of each method.