Alternative splicing is one of the major sources of the large transcriptional diversity found in human cells. Splicing variants have been shown to be associated with features like spreading and progression in several human tumors. Therefore, such variants may be of great importance as both diagnostic and therapeutic tools. Here, by using a set of criteria regarding the expression pattern of splicing variants and statistical analyses, we were able to screen the genome for exons overexpressed in tumors of specific tissues. However, as in other analyses attempting to identify tumor-associated variants, our list of candidates was seriously inflated with cases of genes differentially expressed in tumors. To exclude these cases and increase the probability of finding bona fide regulated splicing variants, we performed a serial analysis of gene expression (SAGE), excluding those genes that were shown to be upregulated in tumors. This allowed us to predict the overexpression of single exons in specific tumors. Our final group of candidates includes 1,386 exons belonging to 638 genes. Experimental validation of a few candidates in normal tissue, tumor cell lines, and patient samples suggests that most of these candidates are indeed tumor-associated exons. Further functional classification of our candidate genes shows that our final list is slightly inflated with cancer-related genes.