Identification of cis-regulatory motifs has long been a hotspot in the study of alternative splicing. We propose a two-step approach: we first identify k-mer seed motifs by testing for enrichment and significant differences in exon skipping rate, then a local stochastic search is applied to refine the seed motifs. Our approach is especially suitable to discover short and degenerate motifs. We applied our method to a dataset of CNS-specific cassette exons in mouse and discovered 15 motifs. Two of these motifs are highly similar to validated motifs, Nova and hnRNP A1 binding sites. Four motifs show positional bias relative to the splice sites. Our study provides a dictionary of sequence motifs involved in the regulation of alternative splicing in CNS tissues, and a novel tool to detect such motifs.