Alternative splicing has emerged as a mechanism that can account for a large proportion of the disparity between the modest number of genes in the human genome and the much higher complexity of the expressed proteome. At least a third, and probably the majority, of human genes are alternatively spliced, and some genes can generate thousands of protein isoforms by complex alternative splicing events. Analysis of the transcriptome will therefore require the development of massively parallel technologies that are able to encompass the complexity arising from alternative splicing.