Here we present evidence, based on 10 datasets comprising 5283 sequences for 200 genera, that the use of the Kimura-2-parameter (K2P) model in DNA-barcoding studies is poorly justified. We demonstrate that K2P is neither expected nor confirmed to be an appropriate model for closely related COI sequences. In addition, we show that the use of uncorrected distances yields higher or similar identification success rates for neighbour-joining trees and distance-based identification techniques. K2P also does not widen the barcoding gap for closely related sequences. We conclude that the spread of K2P through the barcoding literature is difficult to explain, and urge the use of evidence-based approaches to DNA barcoding.
© The Willi Hennig Society 2011.