We have developed a microarray method, based on genomic representations, designed to measure genome copy number fluctuations. A representation is a reproducible sampling of the genome, prepared by digesting genomic DNA with a restriction enzyme such as BglII, ligation of adaptors and polymerase chain reaction amplification. The representation is reduced in complexity in comparison with the original genomic DNA, resulting in an increase in the ratio of signal to noise on the microarray. The probes on the microarray are fragments cloned from a representation of a normal male, and the samples hybridized are representations of genomic DNA made in the same manner as the arrayed fragments. We had previously shown that representations of genomic DNA were faithful and conserved copy number between a test and a reference sample. We now show that representations can be used in a microarray format for analysis on a genome-wide scale, and that this method is reproducible and quantitative. We have used this method to analyze the genome copy number fluctuations in primary breast cancer. We separated specimens into tumor and normal samples by ploidy, prepared representations, labeled them and hybridized them to microarrays. The results were analyzed to identify regions that have undergone gene copy number fluctuations. We have assessed the frequency with which various regions, such as erbB2, have been affected. We have also used this method in an attempt to identify uncharacterized regions. These regions will be subject to closer scrutiny, to identify genes that may play a role in the promotion of primary breast tumorigenesis.