Avian mobbing: byproduct mutualism not reciprocal altruism.

performance of the methods [3]. There is a place for different approaches to reflect the diversity of data and hypotheses being addressed, but it is important that supertree workers justify their choice of method. MRP remains popular largely because of its convenience (as standard phylogenetic software is used to analyse the transformed data), but relying exclusively upon MRP is increasingly difficult to defend. By proposing a statistical model for the supertree problem , the maximum-likelihood approach is a step in the right direction. Bringing phylogenetic supertrees into the same framework as phylogenetic inference from sequence data should help users focus on the importance of meth-odological issues, as statistical model selection is now routine when building phylogenies from sequences. In combination with increasingly realistic models of sequence evolution, maximum-likelihood approaches to supertree construction look set to play an important role in the growing field of phylogenomics. Much work remains to be done: it is unclear how well the mathematically convenient exponential distribution models incongruence in real data, and work has only just begun on efficiently estimating optimal trees under this model. Whether or not this particular approach is successful, supertree methods might need to become part of the mainstream toolkit of molecular phylogenetics if systematists are to make proper use of the deluge of genomic sequence data.