Current strategies for hepatic gene therapy are either quantitatively inefficient or suffer from lack of permanent gene expression. We have utilized an animal model of hereditary tyrosinaemia type I (HT1), a recessive liver disease caused by deficiency of fumarylacetoacetate hydrolase (FAH), to determine whether in vivo selection of corrected hepatocytes could improve the efficiency of liver gene transfer. As few as 1,000 transplanted wild-type hepatocytes were able to repopulate mutant liver, demonstrating their strong competitive growth advantage. Mutant hepatocytes corrected in situ by retroviral gene transfer were also positively selected. In mutant animals treated by multiple retrovirus injections >90% of hepatocytes became FAH positive and liver function was restored to normal. Our results demonstrate that in vivo selection is a useful strategy for hepatic gene therapy and may lead to effective treatment of human HT1 by retroviral gene transfer.