Is the tree of life the best metaphor, model, or heuristic for phylogenetics?

Mindell (2013) [henceforth simply “Mindell”] has provided a claim that the Tree of Life (ToL) is still useful in phylogenetics as a model, a metaphor, and a heuristic. Here I examine all three of these claims. Mindell noted that what biologists have long discussed as the ToL is in fact reticulate, and always has been. He therefore objected to simplistic declarations that there is no ToL, in the sense of no recoverable phylogenetic history for recognized taxa. Mindell argued that a “tree with reticulations” is still basically a tree, and that the ToL therefore continues to be a useful metaphor, model, and heuristic for phylogenetics. Here, I argue an alternative point of view, in which a “tree with reticulations” is a network, and that therefore a network will be a better metaphor, model, and heuristic for phylogenetics, in the sense that it will be more inclusive and more powerful. This distinction between tree and network in the face of reticulations is not a semantic one. The tree metaphor/model/heuristic pre-supposes tree-like data, whereas the network allows the data to determine the tree-likeness of the metaphor/model/heuristic—some networks are more tree-like than are others. So, the network view does not deny the importance of the ToL, but simply makes it a special case of something much more general. In this Point of View I will point out, first, that the tree as a metaphor is actually pre-dated by the network as a phylogenetic metaphor, and that many of the current debates in phylogenetics are not new and actually date back to the beginnings of phylogenetic analysis. We can learn from this history. Second, I point out that the use of a tree as a model pre-supposes a particular approach to data modeling that has serious limitations for phylogenetic analysis. Modeling from complex to simple, rather than simple to complex, has benefits for the analysis. Finally, I note that, as a heuristic, an unrooted network has many advantages over the use of a rooted tree. Data exploration and display is a sorely under-valued part of phylogenetics. Note that I am not arguing that a tree cannot be used as a model, heuristic, or metaphor for phylogenetics, because it surely can, but I am arguing instead that it is not the best model, heuristic, or metaphor—a network is better. Thus, one does not need “a pluralistic view of the ToL” (Mindell, p. 479) but a rather simplistic viewof a network, instead. Mindell sees the ToL as basically a tree even though he recognizes that parts of it are not treelike; but instead we should see an evolutionary history in which some parts are more tree-like than are others.