Abstract When the human mineralocorticoid receptor (hMR) was cloned and sequenced by the Evans laboratory, its close homology with the human glucocorticoid receptor (hGR) was noted; subsequently, on the basis of its higher sequence similarity to the GR than to the progesterone receptor (PR) or androgen receptor (AR), MR and GR have commonly been considered to share an immediate common ancestral “corticoid receptor.” When, however, homology is determined for the full-length receptor, or helices 3/4/5/12, MR is least like any of the other three receptors; for the ligand binding domain, AR is clearly the least homologous. When relative binding and activation capacity of a broad panel of LXXLL cofactor peptides are compared by mammalian two-hybrid assay, GR and PR show a highly similar profile, interacting with many of the 39 peptides, with MR and AR binding very few, supporting a close functional homology between PR and GR. In addition, recent studies by He et al. on FXXLF binding by GR and PR mutants suggest that these two receptors may be closer to AR than to MR. The sequence and transcriptional activity of the S810L mutant reported by Geller et al. can be interpreted as supporting MR as being closer than GR/PR/AR to estrogen receptor/retinoid X receptor, and thus potentially the first rather than the last branch from a common MR/GR/PR/AR ancestor. On the basis of these studies together, we propose reconsideration of the evolutionary tree for the MR/GR/PR/AR subfamily, with MR closest to the primordial ancestral receptor, GR/PR sharing a common immediate ancestor, and a higher degree of evolutionary drift in the AR ligand binding domain to accommodate C19 rather than C21 steroids as physiological ligands.