The classification of the North American darters (Percidae: Etheostomatinae) has been a subject of historical interest and rigorous debate. Conflicting morphological and molecular phylogenetic analyses have made relationships among species unclear and have hampered attempts at classification. Recently we showed that amplified fragment length polymorphism (AFLP) data provide support for relatively ancient relationships (>30 mya) among darters that are consistent with numerous morphological hypotheses. We predicted that increasing the number of extant species sampled would result in a more accurate estimate of evolutionary relationships and increase the statistical resolution, particularly at deeper nodes, as it would increase the probability that two taxa share homologous alleles (fragments) and link deeper relationships, thus overcoming noise created by homoplasious fragments. To test this prediction, we added 32 species to our original analysis of 69 species and reconstructed evolutionary histories using Bayesian methods for AFLPs to infer phylogenetic relationships. Overall resolution of the tree substantially improved. Bayesian Posterior Probabilities were strong (greater than 0.95) at most of the deeper nodes, and most notably increased to 0.97 and 0.99 BPP at the two most ancient and previously weakly supported nodes in Etheostoma (estimated divergence approximately 26.1–33.5 mya). Our results indicated that dense taxon sampling may be necessary to increase accuracy and resolution when reconstructing ancient evolutionary relationships using AFLPs. Results of the present study were compared with earlier classifications and with recent studies on phylogenetic relationships to identify best-supported genera and subgenera.