It is widely assumed that current phylogenetic methods are fairly accurate at recovering the evolutionary relationships among different species, but evaluating the relative success of this enterprise is a difficult task. This study addresses some fundamental questions associated with generating phylogenetic trees. The complete genomes of five species of Steinernema were sequenced and assembled. Genes were predicted in AUGUSTUS and orthologous genes were found from those data using OrthoMCL. I aligned 3890 genes in MAFFT and eliminated poorly aligned positions with GBlocks. I created individual trees for each gene as well as a supermatrix tree in PAUP*, using a closely related taxon from another genus, Panagrellus redivivus. In the resulting gene trees, I found only a small subset of all the possible topologies. I discovered that the supermatrix tree has the same topology as the topology with the most gene trees in the gene-topology distribution. There are only a small number of histories for all of the genes and many of the genes have the same lineage. I bootstrapped the gene-topology distribution and found that the best-supported topology was sampled 22.1% of the time. I show that many genes must be sampled in order to converge on the topology with the most support from the gene trees in this dataset.