Microarray-based copy number analysis has found its way into routine clinical practice, predominantly for the diagnosis of patients with unexplained mental retardation. However, the clinical interpretation of submicroscopic copy number variants (CNVs) is complicated by the fact that many CNVs are also present in the general population. Here we introduce and discuss a workflow that can be used in routine diagnostics to assess the clinical significance of the CNVs identified. We applied this scheme to our cohort of 386 individuals with unexplained mental retardation tested using a genome-wide tiling-resolution DNA microarray and to 978 additional patients with mental retardation reported in 15 genome-wide microarray studies extracted from the literature. In our cohort of 386 patients we identified 25 clinically significant copy number losses (median size 2.6 Mb), nine copy number gains (median size 2.0 Mb), and one mosaic numerical chromosome aberration. Accordingly, the overall diagnostic yield of clinically significant CNVs was 9.1%. Taken together, our cohort and the patients described in the literature include a total of 1,364 analyses of DNA copy number in which a total of 11.2% (71.9% losses, 19.6% gains, 8.5% complex) could be identified, reflecting the overall diagnostic yield of clinically significant CNVs in individuals with unexplained mental retardation.