A statistical approach to constructing cladograms that seeks to find those trees that are most likely to explain the observed data. An explicit model is required.