The method of covarions, or concomitantly variable codons, is a technique in computational phylogenetics that allows the hypothesized rate of molecular evolution at individual codons in a set of nucleotide sequences to vary in an autocorrelated manner.
Under the covarion model, the rates of evolution on different branches of a hypothesized phylogenetic tree vary in an autocorrelated way, and the rates of evolution at different codon sites in an aligned set of DNA or RNA sequences vary in a separate but autocorrelated manner.
Developing a computational algorithm suitable for identifying sites with high evolutionary rates from a static dataset is a challenge due to the constraints of autocorrelation.
The original statement of the method used a rough stochastic model of the evolutionary process designed to identify transiently high-variability codon sites.
Abandoning the requirement that rates be autocorrelated on a given DNA or RNA molecule allows extension of substitution matrix methods to the covarion model.