Data from: Multiple regression modelling for estimating endocranial volume in extinct Mammalia

The profound evolutionary success of mammals has been linked to behavioral
and life-history traits, many of which have been tied to brain size.
However, studies of the evolution of this key trait have yet to explore
the full potential of the fossil record, being limited by the difficulty
of obtaining endocranial data from fossils. Using measurements of
endocranial volume, length, height, and width of the braincase in 503
adult specimens from 199 extant species, representing 99 of 133 extant
mammalian families, we expand upon a simple method of using multiple
regression to develop a formula for estimating brain size from external
skull measurements. We also examined non- mammalian synapsids to assess
the phylogenetic limits of our model's application. Model-predicted
volume correlates strongly with measured volume (R2 = 0.993) and
prediction error is between 16% and 19%. Error decreases if models
developed for well-sampled subclades such as primates or rodents are used,
demonstrating that some differential evolution of the relationship between
brain size and skull size has occurred. However, reanalysis using
phylogenetically independent contrasts demonstrates weak phylogenetic
dependency, indicating that our model is appropriate for estimating the
endocranial volume of species of unknown phylogenetic affinity. Thus, the
model represents a generally applicable, fast and cost-efficient way to
dramatically expand the taxonomic and temporal scope of mammalian brain
size data sets. Even endocranial volumes of taxa with highly derived
crania, such as cetaceans and monotremes, can be estimated confidently.
However, the model works best for generalized placental crania.
Fundamental differences in cranial architecture suggest that the model
cannot provide accurate estimates of endocranial volume in non-mammalian
synapsids more basal than Morganucodon (ca. 200 Ma). Therefore, use of the
model for taxa phylogenetically distant from the mammalian crown group is
not warranted, but it might be used to establish relative brain sizes
between closely related subgroups.