Spinal muscular atrophy (SMA) is characterized by degeneration of motor neurons in the spinal cord, causing progressive weakness of the limbs and trunk, followed by muscle atrophy. SMA is one of the most frequent autosomal recessive diseases, with a carrier frequency of 1 in 50 and the most common genetic cause of childhood mortality. The phenotype is extremely variable, and patients have been classified in type I–III SMA based on age at onset and clinical course. All three types of SMA are caused by mutations in the survival motor neuron gene (SMN1). There are two almost identical copies, SMN1 and SMN2, present on chromosome 5q13. Only homozygous absence of SMN1 is responsible for SMA, while homozygous absence of SMN2, found in about 5% of controls, has no clinical phenotype. Ninety‐six percent of SMA patients display mutations in SMN1, while 4% are unlinked to 5q13. Of the 5q13‐linked SMA patients, 96.4% show homozygous absence of SMN1 exons 7 and 8 or exon 7 only, whereas 3.6% present a compound heterozygosity with a subtle mutation on one chromosome and a deletion/gene conversion on the other chromosome. Among the 23 different subtle mutations described so far, the Y272C missense mutation is the most frequent one, at 20%. Given this uniform mutation spectrum, direct molecular genetic testing is an easy and rapid analysis for most of the SMA patients. Direct testing of heterozygotes, while not trivial, is compromised by the presence of two SMN1 copies per chromosome in about 4% of individuals. The number of SMN2 copies modulates the SMA phenotype. Nevertheless, it should not be used for prediction of severity of the SMA. Hum Mutat 15:228–237, 2000. © 2000 Wiley‐Liss, Inc.