Original Paper Brain Behav Evol 2004;63:47–60 Received: October 8, 2002 Returned for revision: January 22, 2003 DOI: 10.1159/000073759 Accepted after revision: August 8, 2003 Brain Size Scaling and Body Composition in Mammals P. Thomas Schoenemann Department of Anthropology, University of Pennsylvania, Philadelphia, Pa., USA Key Words mass is a more appropriate scaling parameter for com- Mammals W Encephalization W Fat-free weight W Brain paring brain size across species than is overall body size W Body composition W Neanderthal W Sex differences weight. Copyright © 2004 S. Karger AG, Basel Abstract Brain size scales with body size across large groups of Introduction animals, but exactly why this should be the case has not been resolved. It is generally assumed that body size is a It has long been known that brain weight scales with general proxy for some more important or specific un- body weight across large groups of animals. The correla- derlying variable, such as metabolic resources available, tion between these variables in different studies is very surface area of the body, or total muscle mass (which is high, typically greater than r = 0.95 [Martin, 1981]. Exact- more extensively innervated than is, e.g., adipose tis- ly why this would be the case remains a fundamental sue). The present study tests whether brain size in mam- question in the study of brain evolution. Although exis- mals scales more closely with muscle mass (and other tence of a correlation does not prove direct causation, any components of lean body mass) than with total fat. Fel- truly causal relationship must necessarily also be a corre- senstein’s independent comparisons method was used lational one. To the extent we accept that a given correla- to control for phylogenetic effects on body composition tion is not due to chance, we also accept that some causal in organ weight data taken from a previously published connection exists between the two variables, although the comparative sample of 39 species in 8 different orders of connection might only be through the influence of other mammals, all collected and processed by the same (perhaps unmeasured) variables. A key step in under- researchers. The analysis shows that the size of the cen- standing why two variables are correlated involves explor- tral nervous system (CNS) is more closely associated ing the web of associations among possibly relevant vari- with components of fat-free weight than it is to fat ables [Blalock, 1970]. In a 1990 review article in Science, weight. These results suggest a possible explanation for Harvey and Krebs point out that in studies of brain/body why metabolic resources and brain size both share the size scaling, ‘... for the most part, body weight is employed same general relationship with body size across mam- as a surrogate measure for some (perhaps unidentified) mals. They also suggest that some measure of lean body underlying variable.’ [Harvey and Krebs, 1990]. Body © 2004 S. Karger AG, Basel P. Thomas Schoenemann ABC 0006–8977/04/0631–0047$21.00/0 Department of Anthropology, University of Pennsylvania Fax + 41 61 306 12 34 Philadelphia, PA 19104 (USA) E-Mail [email protected] Accessible online at: Tel. +1 215 573 7671, Fax +1 215 898 7462 www.karger.com www.karger.com/bbe E-Mail [email protected] weight might simply be associated with brain weight (in sue. In fact, this argument has long been used as a possible the statistical sense) without directly causing differences explanation for the larger average brain size (as estimated in brain weight. Many biological features covary with by cranial capacity) found in Neanderthal specimens body weight, such that the exact nature of the cause-effect compared to modern humans [Dubois, 1921; Trinkaus relationship between body weight and brain weight re- and Howells, 1979; Holloway, 1981, 1985]. Neanderthal mains unclear. post-crania suggest they were significantly more heavily If body weight is just a proxy measure, what is it a muscled than modern humans [Trinkaus and Howells, proxy for? Two well known hypotheses concern body sur- 1979; Ruff et al., 1993, 1994; Abbott et al., 1996; Trink- face area [Jerison, 1973] and metabolic resources [Martin, aus, 1997]. If brain size varies specifically with muscle 1981; Armstrong, 1983]. Jerison’s [1973] data suggested mass (or something closely associated), we would expect brain and body weight scaled with an allometric exponent Neanderthals to have larger brains. In fact, at least one of F0.67, which is the same as that relating surface area recent study [Wood and Collard, 1999] suggests that the to volume (among identically shaped objects). He there- encephalization quotient (EQ: the ratio of actual/pre- fore suggested that brain size might be keeping pace with dicted brain size) for Neanderthal is essentially the same the surface area of the body, perhaps because the amount as in modern humans. Although there are problems with of information about the external world that a species has this analysis (EQ’s hide large absolute differences in brain access to (and needs to process) might be proportional to size; see discussion below), it supports the contention that surface area [Jerison, 1985]. Martin [1981] used a larger muscle and/or lean body mass might be highly relevant dataset and a more statistically appropriate line-fitting for making sense of differences in estimated cranial technique (major axis) and found that the allometric capacities in hominids. exponent was actually closer to 0.76, which matches In addition, the idea that brain weight might scale spe- empirical estimates of the exponent relating basal meta- cifically with fat-free weight has also long been used as a bolic rate to body size. From this he hypothesized that the possible explanation for sex differences in brain size mother’s basal metabolic rate might be the key limiting (which average about 100 g favoring males, after correct- factor for the size of the brain of any offspring. This argu- ing for differences in total body weight) [Manouvrier, ment has been criticized on the basis that the association 1903; Gould, 1981; Ankney, 1992]. Sexual dimorphism in between brain and metabolism disappears after control- fat-free weight is substantially greater than is sexual ling for body size [McNab and Eisenberg, 1989; Pagel and dimorphism in overall weight (males weigh more and Harvey, 1989], although this simply shows that the varia- have proportionately less fat). tion in brain size that is unrelated to body size is also However, because metabolic resources have been im- unrelated to metabolism. Because we are interested spe- plicated in brain size variation and brain evolution [Aiello cifically in the variance that is related to body size, these and Wheeler, 1995], and because adipose tissue repre- criticisms do not disprove the metabolic resources hy- sents an important storage of metabolic resources, it is pothesis. It is likely that the metabolic resources of an possible that brain size would vary closely with fat levels organism are relevant to brain size variation in some fash- across mammals. It is true that muscle also contains sig- ion [Aiello and Wheeler, 1995]. In any case, these studies nificant metabolic stores (primarily in the form of glyco- demonstrate that in order to explain the correlation gen), and is also very metabolically active. However, gly- between brain size and body size, we must fully investi- cogen stores in muscle are generally unavailable to other gate the web of associations among various biological fea- tissues [Lehninger, 1982], and in any case such stores are tures in groups of mammals and other animals. more limited compared to fat. The body typically has only The present study seeks to expand our understanding enough glycogen to supply about a day’s worth of basal of these associations by focusing on sub-components of metabolism, whereas fat stores generally contain enough body size. Body size is, after all, the sum of a number of calories for several weeks [Lehninger, 1982; Campbell, different parts. These parts are clearly inter-related, but 1993]. From a metabolic perspective, fat would seem to they are not perfect functions of each other. Some sub- be a particularly important gauge of a species energy components might be more closely related to brain size resources. However, fat and muscle are obviously not in than others. Given that muscle tissue is more intimately competition with each other: fat stores exist to supply a associated with brain function than adipose tissue, it is variety of energy needs in the body, including those of reasonable to expect that some estimate of muscle mass muscles. So it is reasonable to expect that muscle mass would scale more closely with brain size than adipose tis- and fat mass would be associated with one another across 48 Brain Behav Evol 2004;63:47–60 Schoenemann species, and therefore that fat and muscle might both scale MacLarnon, 1996], the addition of spinal cord weight should not closely with brain size if metabolic resources are critical. materially affect the conclusions and analyses presented here. It remains an open question, therefore, whether brain Table 1 lists the species used, along with the location and num- bers of individuals of each sex, and the raw data for these measures as weight should be more closely associated with fat-free calculated from tables in Pitts and Bullard [1968]. Five of the eight weight, fat weight, or both. This question has never been orders of mammals are represented by only one species each: Marsu- empirically investigated on any comparative dataset to pials, Edentates, Primates, Insectivores, and Lagomorphs. There date, even though the assumption of tight scaling with were six Carnivore species, nine Chiroptera, and 19 Rodent.