To assess the clinical utility of genome-wide oligonucleotide arrays in diagnosis of mental retardation and to address issues relating to interpretation of copy number changes (CNCs), we collected results on a total of 1499 proband patients from five academic diagnostic laboratories where the same 44K array platform has been used. Three of the five laboratories achieved a diagnostic yield of 14% and the other two had a yield of 11 and 7%, respectively. Approximately 80% of the abnormal cases had a single segment deletion or duplication, whereas the remaining 20% had a compound genomic imbalance involving two or more DNA segments. Deletion of 16p11.2 is a common microdeletion syndrome associated with mental retardation. We classified pathogenic CNCs into six groups according to the structural changes. Our data have demonstrated that the 44K platform provides a reasonable resolution for clinical use and a size of 300 kb can be used as a practical cutoff for further investigations of the clinical relevance of a CNC detected with this platform. We have discussed in depth the issues associated with the clinical use of array CGH and provided guidance for interpretation, reporting, and counseling of test results based on our experience.