The interpretation of hidden support in combined data phylogenetics

In phylogenetic analysis, support for a given clade is hidden when isolated partitions support that clade less than in the analysis of combined data sets. In such simultaneous analyses, signal common to the majority of partitions dominates the topology at the expense of any signal idiosyncratic to each partition. This process is often referred to as synergy and is commonly used to validate the combination of disparate data partitions. We investigate the behaviour of hidden branch support (HBS), partitioned branch support (PBS) and hidden synapomorphy (HS) as measures of hidden support using artificial, real and experimentally manipulated phylogenetic data sets. Our analyses demonstrate that high levels of both HBS and HS can be obtained by combining data with little shared phylogenetic signal. This finding is in agreement with the original intent of hidden support metrics, which essentially quantify the extent of data set interaction, both through the dispersion of homoplasy and revelation of underlying shared signal (positive data synergy). High levels of HBS alone are insufficient to justify data combination. We advocate the use of multiple hidden support measures to distinguish between the dispersion of homoplasy and positive data synergy, and to better interpret data interactions. Furthermore, we suggest two criteria that help identify hidden support resulting from homoplasy dispersion: first, when total support decreases with the addition of a data partition and second, when total HBS per unit total support (TS) per node is similar to that derived from randomized data.