Mental retardation affects 2 to 3% of the population and is marked by significant etiological heterogeneity, including genetic and non genetic causes. FRAXA (FMR1) trinucleotide expansion is widely searched in routine screening, but found in only about 2% of the patients tested. Mutations of the MECP2 (methyl-CpG-binding protein) gene mainly cause Rett syndrome but were also shown to be involved in mental retardation. This study aimed to estimate the frequency of MECP2 gene mutations in a large group of mentally retarded patients without FRAXA expansion. Screening by heteroduplex analysis and SSCP followed by DNA sequencing of shifted bands were performed on 613 patients, including 442 males and 171 females. Eleven sequence variants were found, including nine polymorphisms. The two others may be pathogenetic. The first one, the double nucleotide substitution c.1162_1163delinsTA leading to a premature stop codon (p.Pro388X) was found in a female patient with random X-inactivation, presenting with borderline mental impairment without any features of Rett syndrome. The second one, the c.679C>G substitution, changing a glutamine to a glutamate in the transcriptional repression functional domain (p.Gln227Glu), was found in a female patient with a moderately biased X-chromosome inactivation profile and presenting with mild intellectual delay and minor psychotic features. The low mutation rate suggests that a large-scale routine screening for MECP2 in mentally retarded subjects is not cost-effective in clinical practice. Screening may be improved by a pre-selection based on clinical features that remain to be established.