Maximisation Principles in Evolutionary Biology 345Nowadays human geneticists implicitly use the same argument to infer relation-
ship given the mode of inheritance: if a rare gene is found in two people in a small
community the instant reaction is to enquire whether they are related.
8 RECONSTRUCTING EVOLUTION: THE METHOD OF MINIMUM
EVOLUTIONIn 1963 Cavalli-Sforza and Edwards suggested the following principle for estimating
the shape and dimensions of a phylogenetic tree:
THE PRINCIPLE OF MINIMUM EVOLUTION
The most plausible estimate of the evolutionary tree is that which
invokes the minimum nett [sic] amount of evolution.They thus made explicit what had for so long been implicit in phylogenetic
thinking, and they did so because the existence of the new electronic computers
which they were harnessing to the problem compelled the replacement of general
ideas like Ockham’s razor and the Darwin principle with unambiguous concepts
which wereprogrammable. As soon as objects that are to be connected by a
phylogenetic tree can be displayed in a multi-dimensional character-space (it need
not even be Euclidean — with DNA sequences, for example, it is a lattice; [Edwards
and Cavalli-Sforza, 1964]) the construction of a tree connecting them with the
minimum total edge-length is a programmable problem. The first example of a
solution was presented at the International Congress of Genetics in 1963 [Cavalli-
Sforza and Edwards, 1964].
Cavalli-Sforza and Edwards were quick to claim that the justification for their
‘method of minimum evolution’, as it became known, was statistical, based on a
probability model for the evolutionary process coupled with standard statistical
estimation of the tree. Having originally used the method of least squares, they
soon developed a detailed probability model which enabled them to apply the
method of maximum likelihood. They then underpinned their method of minimum
evolution with the argument that it gave results which were close to the maximum-
likelihood solution, thus again reflecting the link between traditional phylogenetic
procedures and their justification in terms of likelihood. The history has been given
by Edwards [1996], while the bookInferring Phylogeniesby Felsenstein [2004] is a
rich source of information about the many statistical procedures which have been
developed in the past forty years.
Unfortunately some of the terminology has changed in the course of time. Camin
and Sokal [1965] introduced the phrase ‘principle of evolutionary parsimony’ but
it inevitably became shortened, in the phylogenetic context, to ‘principle of parsi-
mony’, thus confusing it with Ockham’s razor, while ‘method of minimum evolu-
tion’ was used in a new (and now dominant) sense by Rzhetsky and Nei [1993].
The three common procedures currently in use, minimisation of the total tree
length (‘parsimony’), least squares applied to a matrix of pairwise distances, and