Spinal muscular atrophy (SMA), an autosomal recessive neuromuscular disorder, is the leading genetic cause of infant mortality. SMA is caused by the homozygous loss of Survival Motor Neuron-1 (SMN1). In humans, a nearly identical copy gene is present, SMN2. SMN2 is retained in all SMA patients and encodes the same protein as SMN1. However, SMN1 and SMN2 differ by a silent C-to-T transition at the 5' end of exon 7, causing alternative splicing of SMN2 transcripts and low levels of full-length SMN. SMA is monogenic and therefore well suited for gene-replacement strategies. Recently, self-complementary adeno-associated virus (scAAV) vectors have been used to deliver the SMN cDNA to an animal model of disease, the SMND7 mouse. In this study, we examine a severe model of SMA, Smn -/-;SMN2 + / + , to determine whether gene replacement is viable in a model in which disease development begins in utero. Using two delivery paradigms, intracerebroventricular injections and intravenous injections, we delivered scAAV9-SMN and demonstrated a two to four fold increase in survival, in addition to improving many of the phenotypic parameters of the model. This represents the longest extension in survival for this severe model for any therapeutic intervention and suggests that postsymptomatic treatment of SMA may lead to significant improvement of disease severity.