Relative Rates of Nucleotide Substitution. Early work on chloroplast sequences suggested that substitution rates do not follow a constant molecular clock. Rodermel and Bogorad (1987) first claimed that substitution rates in chloroplast loci can vary among evolutionary lineages. They found that the atpE gene had slower missense rates, and the atpH gene had slower synonymous rates, in grass lineages relative to dicot lineages. Similar comparisons of substitution rates in the rbcL locus indicated that both missense and synonymous rates were faster in grass species relative to palm species (Wilson et al., 1990).

These studies relied on fossil-based divergence times for substitution rate estimates. Fossil-based divergence times can have large errors that introduce large (and unmeasurable) uncertainty into rate estimates. Relative rate tests allow comparison of substitution rates between evolutionary lineages without dependence on knowledge of the time dimension. First utilized by Sarich and Wilson (1973), Wu and Li (1985) later extended relative rate tests to nucleotide sequences using a distance measure formulation. Subsequent refinements include a maximum likelihood relative test (Muse and Weir, 1992) and a simple counting method (Tajima, 1993). The power of these methods to detect deviations from a molecular clock depends on a number of factors (e.g., the number of substitution events in the lineages and the transition/transversion bias), but in many situations the three methods have comparable power (Muse and Weir, 1992; Tajima, 1993). Relative rates tests cannot detect changes in evolutionary rates if rates change proportionally among lineages (Fitch, 1976), although this precise condition seems unlikely to occur often. Relative rate tests may also fail to detect stochastic changes in rate within a lineage because the test compares average substitution rates. More importantly, relative rates contain less information on variation than do absolute rates, but absolute rates are difficult to estimate because of imperfect knowledge of the time dimension.

The large rbcL data base has facilitated the characterization of substitution rate variation in a number of evolutionary lineages. A few studies have applied relative rate tests to rbcL sequences from intrafamilial taxa (Soltis et al., 1990; Doebley et al., 1990; Bousquet et al., 1992a); on the whole, these studies have uncovered limited rate variation. Two studies have characterized rbcL rate variation over a wider sampling of plant taxa (Bousquet et al., 1992b; Gaut et al., 1992). Gaut et al. (1992) examined rbcL sequences from 35 monocotyledonous taxa and found substantial synonymous rate variation among evolutionary lineages, but little missense rate variation. The analyses revealed rate homogeneity for rbcL sequences within well-defined families, but substantial rate heterogeneity in interfamilial contrasts. The pattern of interfamilial rate variation