The etymology and historic usage of such terms as ‘anabranch’, ‘anastamose’ and ‘braided’ within river science are reviewed. Despite several decades of modern research to define river channel typologies inclusive of single channels and multiple channel networks, typologies remain ill-conditioned and consequently ill-defined. Conventionally employed quantitative planform characteristics of river networks possibly cannot be used alone to define channel types, yet the planform remains a central part of all modern classification schemes, supplemented by sedimentological and other qualitative channel characteristics. Planform characteristics largely have been defined using non-standardized metrics describing individual network components, such as link lengths, braiding intensity and bifurcation angles, which often fail to separate visually-different networks of channels. We find that existing typologies remain pragmatically utilitarian rather than fundamentally physics-based and too often fail to discriminate between two distinctive and important processes integral to new channel initiation and flow-splitting: (i) in-channel bar accretion, and (ii) channel avulsion and floodplain excision. It is suggested that, first, if channel planform is to remain central to river typologies, then more rigorous quantitative approaches to the analysis of extended integral channel networks at extended reach scales (rather than network components) are required to correctly determine whether ‘visually-different’ channel patterns can be discriminated consistently; and, second, if such visually-different styles do in fact differ in their governing processes of formation and maintenance. A significant question is why do so many seemingly equilibrium network geometries possess a large number of anabranches in excess of predictions from theoretical considerations? The key research frontier with respect to initiating and maintaining multichannel networks remains the understanding and discrimination of accretionary-bar flow splitting versus avulsive processes. Existing and new knowledge on flow splitting processes needs to be better integrated into channel typologies. Copyright © 2013 John Wiley & Sons, Ltd.