Spinal muscular atrophy (SMA) is an autosomal recessive neuromuscular disorder caused by absence of or mutations in the survival motor neuron1 gene (SMN1). All SMA patients have a highly homologous copy of SMN1, the SMN2 gene. Severe (type I) SMA patients present one or two SMN2 copies, whereas milder chronic forms (type II-III) usually have three or four SMN2 copies. SMN2 dosage is important to stratify patients for motor function tests and clinical trials. Our aim was to compare three methods, marker analysis, real-time quantitative polymerase chain reaction using the LightCycler instrument, and multiple ligation-dependent probe amplification (MLPA), to characterize their accuracy in quantifying SMN2 genes. We studied a group of 62 genetically confirmed SMA patients, 54 with homozygous absence of exons 7 and 8 of SMN1 and 8 with SMN2-SMN1 hybrid genes. A complete correlation using the three methods was observed in 32 patients (51.6%). In the remaining 30 patients, discordances between the three methods were found, including under or overestimation of SMN2 copies by marker analysis with respect to the quantitative methods (LightCycler and MLPA) because of lack of informativeness of markers, 3' deletions of SMN genes, and breakpoints in SMN2-SMN1 hybrid genes. The technical limitations and advantages and disadvantages of these methods are discussed. We conclude that the three methods complement each other in estimating the SMN2 copy number in most cases. However, MLPA offers additional information to characterize SMA cases with particular rearrangements such as partial deletions and hybrid genes.