UNINFORMATIVE BOOTSTRAPPING

Abstract
The effect of uninformative characters on the "significance levels" obtained by bootstrapping in cladistic analysis is investigated empirically. Twenty-eight data sets from Platnick's benchmarks are analysed with Siddall's Random Cladistics bootstrapping module and Farris'Hennig86 program, using 1000 replicates and exact calculations. Contrary to the assurances of Harshman, inclusion of autapomorphies in a matrix commonly leads to a loss of "significance".