Early efforts to reconstruct the course of the evolution of the human brain relied on comparing the brains of a few related mammals with brains at successively higher levels of complexity. This Clark or ladder of levels approach is now seen as having limited usefulness in that species are not easily assigned to levels, and extant mammals are now recognized as mosaics of primitive and derived features. In addition, direction of change does not necessarily proceed from simple to complex, small to large, or diffuse to differentiated. A modern cladistic approach reconstructs the brains of ancestors by identifying brain characters within and across phylogenetic groups (clades), and uses parsimony or likelihood to infer direction of change and distinguish ancestral features from independently evolved convergences. Unfortunately, an idealized cladistic approach is often difficult to realize because characters may be hard to identify and validate, key species may be unavailable for study, and broadly based comparative studies can be costly, poorly funded, and labor intensive. Thus, many investigators pursue a truncated approach that is superficially Clark-like but conceptually cladistic. A truncated approach that relies on the extensive study of a few species may compensate for weaknesses by including niche-matched species that offer the opportunity to estimate the likelihood of similar brain features evolving as convergent adaptations. Because inferences about the brains of the primate ancestor are often made from the brains of tree shrews, we compare the brains of squirrel-like tree shrews with the brains of diurnal squirrels, and suggest that many of the primate-like features of the visual system of tree shrews arose independently of those in primates.