Loss of selective constraint on a gene may be expected following changes in the environment or life history that render its function unnecessary. The long-term persistence of protein-coding genes after the loss of known functional necessity can occur by chance or because of selective maintenance of an unknown gene function. The selective maintenance of an alternative gene function is not demonstrated by the failure of statistical tests to reject the hypothesis that there has been no change in the degree of constraint on the evolution of coding genes. Maintenance may be inferred, however, when power analyses of such tests demonstrate that there has been a sufficient number of nucleotide substitutions to detect the loss of selective constraint. Here, we describe a power analysis for tests of loss of constraint on protein-coding genes. The power analysis was applied to loss-of-constraint tests for opsin gene evolution in cave-dwelling crayfish and rbcL evolution in nonphotosynthetic parasitic plants. The power of previously applied tests for loss of constraint on cave crayfish opsin genes was insufficient to distinguish between chance retention and selective maintenance of opsin genes. However, the power of codon-based likelihood ratio tests for change in d(N)/d(S) (=omega) (nonsynonymous to synonymous change) did have sufficient power to detect a loss of constraint on rbcL associated with a loss of photosynthesis in most examples but failed to detect such a change in three independent lineages. We conclude that rbcL has been selectively maintained in these holoparasitic plant lineages. This conclusion suggests that either these taxa are photosynthetic for at least a part of their life or rbcL may have an unknown function in these plants unrelated to photosynthesis.