An analysis of the Kimura 3ST model of DNA sequence evolution is given on the basis of its continuous Lie symmetries. The rate matrix commutes with a U(1) × U(1) × U(1) phase subgroup of the group GL(4) of 4 × 4 invertible complex matrices acting on a linear space spanned by the four nucleic acid base letters. The diagonal 'branching operator' representing speciation is defined, and shown to intertwine the U(1)×U(1)×U(1) action. Using the intertwining property, a general formula for the probability density on the leaves of a binary tree under the Kimura model is derived, which is shown to be equivalent to established phylogenetic spectral transform methods.The use of Markov models of stochastic change to taxonomic character distributions is part of the standard armoury of techniques for describing mutations and inferring ancestral relationships between taxa. For the simplest models, symmetries of the rate matrix under discrete group actions (Z 2 for binary types, or Z 2 × Z 2 for DNA or RNA bases in molecular applications, for example) have been used to good effect in simplifying phylogenetic analysis (for references, see below). In particular, much attention has been centred on properties of the frequently used Kimura 3ST model [1] which possesses such symmetry.A general framework for phylogenetic branching models is as follows [2]. By assumption, different taxonomic units are identified, and classified by a set of defining characteristics: for example, based on morphological features or on sequence data, say, for a particular gene or protein. To each taxon is ascribed a character probability density, and it is the task of phylogenetic reconstruction to infer ancestral relationships amongst a group of related taxa, given sample character frequencies.In this letter, we describe an approach to the analysis of symmetries of such models using continuous transformation groups. Rather than identifying the character types with