There are no disease-modifying treatments for adult human neurodegenerative diseases. Here we test RNA-targeted therapies(1) in two mouse models of spinocerebellar ataxia type 2 (SCA2), an autosomal dominant polyglutamine disease(2). Both models recreate the progressive adult-onset dysfunction and degeneration of a neuronal network that are seen in patients, including decreased firing frequency of cerebellar Purkinje cells and a decline in motor function(3,4). We developed a potential therapy directed at the ATXN2 gene by screening 152 antisense oligonucleotides (ASOs). The most promising oligonucleotide, ASO7, downregulated ATXN2 mRNA and protein, which resulted in delayed onset of the SCA2 phenotype. After delivery by intracerebroventricular injection to ATXN2-Q127 mice, ASO7 localized to Purkinje cells, reduced cerebellar ATXN2 expression below 75% for more than 10 weeks without microglial activation, and reduced the levels of cerebellar ATXN2. Treatment of symptomatic mice with ASO7 improved motor function compared to saline-treated mice. ASO7 had a similar effect in the BAC-Q72 SCA2 mouse model, and in both mouse models it normalized protein levels of several SCA2-related proteins expressed in Purkinje cells, including Rgs8, Pcp2, Pcp4, Homer3, Cep76 and Fam107b. Notably, the firing frequency of Purkinje cells returned to normal even when treatment was initiated more than 12 weeks after the onset of the motor phenotype in BAC-Q72 mice. These findings support ASOs as a promising approach for treating some human neurodegenerative diseases.