Chromosomal loss in cancer cells has been observed by nonisotopic in situ hybridization using pericentromeric probes. Accurate determination of this loss depends upon efficient hybridization and visualization of all probe signals in a two-dimensional field. Close association of homologous centromeres, reported in various normal and tumor tissues, can complicate evaluation and interpretation, resulting in the overestimation of actual chromosome loss. Using pericentromeric FISH probes for chromosomes 9 (classical and beta-satellite) and 17 (alpha-satellite), as well as 17q-specific phage probes, we tested the frequency of homologous centromere association in normal and malignant prostate tissues and lymphoblastoid cells. We found that the pericentromeric region of both chromosome 9 and 17 associated in prostate tissues, but only chromosome 9 demonstrated association in the lymphoblastoid cells. The association observed for chromosome 9 in both cell types appeared to be limited to the classical satellite III-type of heterochromatic, pericentromeric DNA. Using a single-copy probe along with the chromosome 17-specific pericentromeric probe, we determined that the association did not extend to the chromosome arms, but was limited to the pericentromeric region of chromosome 17. To accurately determine chromosome loss, we advocate the use of single-copy probes in addition to, or in place of, pericentromeric probes.