Duplicate gene rooting may be especially useful in cases where there is no outgroup available (e.g., the entire tree of life), or in cases where the branch connecting an outgroup and ingroup is very long. In the latter case, one advantage lies in the potential of reducing long branch effects. If a gene were to have duplicated along a long branch leading to an ingroup, the resulting forms of the gene would share a more recent common ancestor than either one would with the single gene present in outgroup taxa, effectively bisecting the long branch. Two approaches to duplicate gene rooting have been elucidated. Under the "reciprocal outgroups" interpretation of the approach, trees from one gene copy are rooted with a single gene or few copies of the other form. Congruence of the separate gene trees is evaluated after separate analyses and provides the criterion for choosing a rooted species tree; if the separate gene trees conflict regarding the root, a decision is not possible. Under the "minimum events" interpretation of the approach, simultaneous analysis of both genes results in an unrooted gene network. Costs of fitting the two-gene network into a set of possible species trees provide the criteria for choosing the best rooted species tree. We discuss the merits of the two approaches, as well as the effects of including outgroups in duplicate gene analyses. Finally, we discuss the potential of the approach to resolve difficult rooting problems in plant phylogeny, and we review the examples of duplicate gene rooting in plant phylogenetic studies to date, including examples using duplicate genes encoding phosphoglucose isomerases, alcohol dehydrogenases, and phytochromes.

Key words: Duplicate genes, rooting