Proportional versus relative size as metrics in human brain evolution

In addressing the question of whether human prefrontal cortex (PFC) is “disproportionately larger in humans compared with nonhuman primates” in their article in PNAS, Donahue et al. (1) gloss over the distinction between proportional and relative size. Agreeing with previous work (2–4), they show that PFC occupies a larger proportion of the cerebral cortex in humans than in chimpanzees and a larger proportion in chimpanzees than in macaques. Contrary to their presentation of the debate, this fact is uncontroversial. However, proportional size is often a misleading functional metric because of nonlinear, or allometric, scaling: A biological structure can differ across species as a proportion of overall size and yet be functionally equivalent (5) (Fig. 1). PFC volume proportion is predictable from overall brain size, even among nonhuman species (2–4). This increase in volume proportions may be a scaling effect, with no consequences for functional influence exerted by PFC if there are different scaling constraints on different cortical regions (6). For example, disproportionate PFC white matter volume and cell size increases may be necessary to maintain long-range neural connectivity and neural transmission speed in larger brains and bodies (4, 6, 7). Indeed, this is consistent with Donahue et al.’s observation that white matter largely explains the species differences they observe in PFC proportional volume.

Fig. 1.

How proportions mislead comparative analyses due to allometric scaling. Allometric scaling results in log–log plots with scaling coefficients >1 (A) or <1 (B). Using proportional size assumes isometry (coefficient = 1, dashed lines). For example, the mammalian skeleton becomes a larger proportion of body mass as size increases (positive allometry) because mechanical stress increases with body mass nonlinearly. Brain mass scales with negative allometry so that smaller mammals have proportionately larger brains. Adaptive variation, after taking scaling laws into account, is reflected in differences in relative size (i.e., the deviation from this general scaling relationship) for individual species (C) or whole clades (D). Data are from refs. 9 and 10.

Donahue et al. (1) “do not view the idea of a predictive positive allometric scaling as mutually exclusive with a preferential expansion of PFC” but provide no justification for this view. Preferential expansion implies the action of natural selection on functions mediated specifically by PFC. Why, then, would selection on PFC function not be reflected in a difference in relative, as well as proportional, size, given that other brain components, including individual cortical regions, can and do show such scaling-independent variability (8)? Donahue et al.’s study of three species prohibits statistical analyses to test this hypothesis, although their figure 3 suggests it is not the case. Hence, their conclusion that human PFC exhibits preferential expansion in humans is based only on comparing the size of PFC as a proportion of total cortical volume.

Conflating variation in proportional size caused by allometry with nonallometric differences in relative size has unpalatable consequences. For example, (i) small species such as mice have larger brains as a proportion of body mass than do humans (Fig. 1), and (ii) frontal gray matter volume is larger as a proportion of cortical volume in several nonhuman primates and carnivores than in humans (4). However, we doubt that either of these observations would lead anyone to ascribe greater cognitive capacities to the nonhuman species.

Instead of applying double standards to metrics according to whether or not they favor humans, we need a more principled approach informed by better understanding of principles of neural allometric scaling. If human PFC increased with other brain structures according to a conserved scaling relationship, it is reasonable to assume that the target of selection was not PFC functions, but functions mediated by more distributed neural systems, and these should therefore be the target of our inquiry (4, 8).