Exon repetition describes the presence of tandemly repeated exons in mRNA in the absence of duplications in the genome. Its existence challenges our understanding of gene expression, because the linear organization of sequences in apparently normal genes must be subverted during RNA synthesis or processing. It is restricted to a small number of genes in some of which over half of the mRNA contains specific patterns of repetition. Although it is sometimes assumed to arise by trans-splicing, there is no evidence of this and the efficiency is very much higher than for examples of bona fide trans-splicing in mammals. Furthermore, a potentially ubiquitous reaction such as trans-splicing is not consistent with a phenomenon that involves such a high proportion of the products of so few genes. Instead, it seems more probable that exon repetition is caused by a specific trans-acting factor. We have tested this and demonstrate for the two best characterized examples that the property is restricted to specific alleles of the affected genes and is determined in cis. It is not determined by exonic splicing signals, as had been suggested previously. In heterozygotes, RNA transcribed from the two alleles of an affected gene can have fundamentally different fates.