have been favored by selection over the canonical way of accomplishing the same task (Speijer, 2007; Ochsenreiter et al., 2008). We find neither of these arguments to be particularly compelling given the narrow distribution of these characters in nature, and their often extreme complexity. For example, dozens of nuclear-encoded proteins are required for T. brucei to edit just 12 mRNAs (Lukeš et al., 2005; Etheridge et al., 2008; Hashimi et al., 2008). Despite considerable controversy, no obvious evolutionary advantage has ever been demonstrated for this type of editing, and such possible advantages that have been proposed (e.g., the generation of 2 proteins from 1 gene) (Ochsenreiter et al., 2008) are more than outweighed by the demonstrated cost (i.e., “save” 1 gene at the cost of dozens of genes). We argue that constructive neutral evolution offers a more compelling explanation (Covello and Gray, 1993; Stoltzfus, 1999). This is a very simple and intuitive way of explaining complexity in biological systems, but one that has not received much attention. Briefly, it is possible for a biological system to increase in complexity (i.e., to increase the number of components or interactions needed to sustain the system) by making a series of neutral changes that collectively do not affect fitness. Pan-editing is often thought of as an error correcting system, but as Stoltzfus (1999) pointed out, the duplicated information (e.g., gRNAs) must have been created before the mutations they are correcting, or they too would carry the mutations—so the error-then-solution model is backward. Instead, if a gratuitous duplication of information took place first (i.e., the origin of a gRNA), then a subsequent mutation could be neutralized by the presence of the duplicated information needed to change it. The fixation of such a mutation would render the gRNA essential, and would also allow for further mutations as long as the gRNAs could mediate their reversal. This last point is important because it would bias the system against the loss of the gRNA since mutations at many sites will further establish the gRNA as essential, whereas only complete reversion to the original sequence could render it unnecessary. Overall, the editing activity and the sites that are edited will coevolve, and the complexity of the system will inevitably grow while conferring no real selective advantage (for many other case studies and much greater detail) [see Covello and Gray (1993) and Stoltzfus (1999)].

Within this framework, together with the recognition that the evolution of an unusual character can be an intrinsic factor in the subsequent evolution of additional, specific characters, a complex cellular system may be explained simply by identifying the event(s) that preconditioned the cell for such a system. Convergence may offer a glimpse into these conditions by revealing how characters are linked when the same events are played out multiple times.