Septin family proteins are quite similar to each other both within and between eukaryotic species. Typically, multiple discrete septins co-assemble into linear heterooligomers (usually hexameric or octameric rods) with a variety of cellular functions. We know little about how incorporation of different septins confers different properties to such complexes. This issue is especially acute in human cells where 13 separate septin gene products (often produced in multiple forms arising from alternative start codons and differential splicing) are expressed in a tissue-specific manner. Based on sequence alignments and phylogenetic criteria, human septins fall into four distinct groups predictive of their interactions, that is, members of the same group appear to occupy the same position within oligomeric septin protomers, which are "palindromic" (have twofold rotational symmetry about a central homodimeric pair). Many such protomers are capable of end-to-end polymerization, generating filaments. Over a decade ago, a study using X-ray crystallography and single-particle electron microscopy deduced the arrangement within recombinant heterohexamers comprising representatives of three human septin groups-SEPT2, SEPT6, and SEPT7. This model greatly influenced subsequent studies of human and other septin complexes, including how incorporating a septin from a fourth group forms heterooctamers, as first observed in budding yeast. Two recent studies, including one in this issue of Cytoskeleton, provide clear evidence that, in fact, the organization of subunits within human septin heterohexamers and heterooctamers is inverted relative to the original model. These findings are discussed here in a broader context, including possible causes for the initial confusion.