Traditionally, vectors for gene transfer/therapy experiments were mono- or bicistronic. In the latter case, vectors express the gene of interest coupled with a marker gene. An increasing demand for more complex polycistronic vectors has arisen in recent years to obtain complex gene transfer/therapy effects. In particular, this demand is stimulated by the hope of a more powerful effect from combined gene therapy than from single gene therapy in a process whose parallels lie in the multi-drug combined therapies for cancer or AIDS. In the 1980's we had only splicing signals and internal promoters to construct such vectors: now a new set of biotechnological tools enables us to design new and more reliable bicistronic and polycistronic vectors. This article focuses on the description and comparison of the strategies for co-expression of two genes in bicistronic vectors, from the oldest to the more recently described: internal promoters, splicing, reinitiation, IRES, self-processing peptides (e.g. foot-and-mouth disease virus 2A), proteolytic cleavable sites (e.g. fusagen) and fusion of genes. I propose a classification of these strategies based upon either the use of multiple transcripts (with transcriptional mechanisms), or single transcripts (using translational/post-translational mechanisms). I also examine the different attempts to utilize these strategies in the construction of polycistronic vectors and the main problems encountered. Several potential uses of these polycistronic vectors, both in basic research and in therapy-focused applications, are discussed. The importance of the study of viral gene expression strategies and the need to transfer this knowledge to vector design is highlighted.