The future of non-viral gene therapy depends on a detailed understanding of the barriers to delivery of polynucleotides. These include physicomechanical barriers, which limit the design of delivery devices, physicochemical barriers that influence self-assembly of colloidal particulate formulations, and biological barriers that compromise delivery of the DNA to its target site. It is important that realistic delivery strategies are adopted for early clinical trials in non-viral gene therapy. In the longer term, it should be possible to improve the efficiency of gene delivery by learning from the attributes which viruses have evolved; attributes that enable translocation of viral components across biological membranes. Assembly of stable, organized virus-like particles will require a higher level of control than current practice. Here, we summarize present knowledge of the biodistribution and cellular interactions of gene delivery systems and consider how improvements in gene delivery will be accomplished in the future.