Brain expansion in A. boisei

Elton and colleagues (2001) examined the record of brain size in early Homo with the following question in mind: we know that brain size increased in this lineage, but was that increase unusual compared to other lineages of primates at the same time? To answer this, they examined the brain sizes in fossil A. boisei and Theropithecus (the genus that includes living gelada baboons). Answering this question would determine whether the brain size of early Homo increased for reasons unique to this genus, or whether instead it was part of a broader trend that might be attributed to climatic changes or other ecological factors.

The results of the study showed that fossil Theropithecus showed no particular trends in brain size over time. But A. boisei did show a significantly positive trend toward brain growth over time. This trend exists whether the early KNM-WT 17000 specimen is included in the sample or not, which is important because this skull is both small and early, and by itself might create a trend in a sample that was otherwise static over time. Without that skull, the trend is still there, driven mainly by the late large skull from Konso KGA 10-525, and the early small juvenile skull Omo L338y-6. Although this latter skull is juvenile, they use an estimated adult size that is about 4 percent larger than the actual endocast.

The study compared these two cases with the evidence for brain size in early Homo. Looking only at Homo habilis, there is no apparent trend toward increasing brain size. This is partly because the largest specimen, KNM-ER 1470, is early and partly because of the great variation within the sample. The overall sample including H. habilis and early humans does show a significant trend over time, but this trend appears mainly to result from the presence of two distinct (and mostly discontiguous) species, one of which survives much later in time and therefore greatly influences the appearance of a trend. Considering early humans alone, there is really no trend evident before 1.5 million years ago, and only a slight increase up to the sample around a million years ago (Lee and Wolpoff 2002).

Some issues:

The study focused on change within each fossil species. But there is no comparison to the magnitude of changes that occurred between hominid taxa. This is problematic because most of the brain evolution in early Homo likely characterized the initial origin of the lineage from an ancestral australopithecine. It is no great surprise that H. habilis does not change markedly over time, but what is surprising is the substantial jump in size from earlier australopithecines like A. afarensis or A. africanus and later Homo. The same could be observed of the change between habilines and early humans. The authors actually run a test to see if the entire early Homo sample shows a trend over time (and it does), but it is clear from the data that the major difference is the shift in size from habilines to early humans, with each of these groups showing relatively little change over time.

The trend in A. boisei depends entirely on the earliest and latest fossils. The small size of the early Omo L338y-6 specimen is unsurprising compared to the even smaller KNM-WT 17000, so the idea that the A. boisei lineage should have changed over time is possibly expected. But Omo L338y-6 is not the smallest member of the later sample (KNM-ER 407 is smaller), so it does make a difference whether KNM-WT 17000 is excluded or not. Especially considering this is a robust probable male skull, its very small endocranial volume makes a large contrast with later A. boisei, a difference extended by many other anatomical details.

What about the late end of the sample? Here, the endocranial volume of KGA 10-525 appears very large, and is at the high end of the A. boisei range. But compared to earlier hominids, the volume is not surprisingly large. For example, the endocranial volume of AL 444-2 (A. afarensis) is estimated at around 550 mL (Holloway and Yuan 2004), and the volume of STW 505 (A. africanus) is certainly larger, perhaps over 600 mL (Hawks and Wolpoff 1999; Conroy et al. 1999). Although the body size of KGA 10-525 is not known, its molars are near the top end of the A. boisei sample, exceeded only by OH 5. This might suggest that the body size of the specimen was among the largest in the sample, and at the least we can guess that the individual was larger than the average for males.

So to address whether KGA 10-525 was surprisingly large, we have to look beyond its date and ask what the expected range of brain sizes within A. boisei would have been. Including KNM-WT 17000 at the small end, and KGA 10-525 at the large end, the standard deviation of the entire A. boisei sensu lato sample in endocranial volume is 39.3 mL. With an average volume of 480 mL, this yields a CV (coefficient of variation) of 8.2 percent.

By contrast, the H. habilis sensu lato sample, including KNM-ER 1470, has a standard deviation of 79.6 mL on an average of 634 mL, yielding a CV of 12.6 percent. So the A. boisei sample is a third less variable than the H. habilis sample.

Holloway (1980) gives CV values for recent humans, from the Danish data of Pakkenberg and Voight (1964), broken down by sex. The within-sex CV's for males and females were 8.2 percent and 8.3 percent, respectively. So the variation within the extant sample of A. boisei, including KNM-WT 17000, is about the same as within one sex in living humans. This is despite the fact that the A. boisei sample spans a million years of time and appears to have been substantially greater in body size dimorphism (as indicated by cranial robusticity and tooth sizes) compared to humans.

Tobias (1971) pools data from several earlier studies of endocranial volumes in hominoids, pooling sexes together. In his summary, the smallest degree of variation is within white-handed gibbons (Hylobates lar), where the CV of endocranial volume is 7.6 percent. Other hominoids are higher: chimpanzees at 9.7 percent, siamangs at 10.7 percent, orangutans at 10.9 percent, and a male-biased sample of gorillas at 13.1 percent. Except for the small and monomorphic gibbons, all these are higher than the estimate for A. boisei.

So the problem is not that KGA 10-525 is surprisingly large. Instead, the problem is that variation in A. boisei has likely been substantially undersampled. There should be many larger and smaller crania than have yet been found in the sample.

This is a problem for testing whether there is a significant trend within the A. boisei sample. In a sample with relatively low variation, the observation of a single large specimen at the recent end of the sample may be statistically surprising--the rarity of the large size is combined with the rarity of the recent date.

In a study of fossils, we cannot really know what the underlying variability of the extinct species was. For this reason, we are left with tests that use only the observed sample variability. The best of these are randomization tests, which randomize one or more elements of the sample to determine the likelihood that the sample would have the observed characteristics based on the data at hand. But randomization tests assume that the data themselves are sufficient to represent the variation in the underlying population. If there is good reason to think that the data are not representative, then the randomization tests may mislead about the chance that the data would be ordered in the observed way at random.

What if instead of randomly ordering the data to test its significance, instead we modeled the characteristics of the underlying population. For example, we could assume that the population had been a single species with a standard deviation similar to that observed in some living or fossil species--perhaps the observed standard deviation for earlier hominids, or for recent humans. The null hypothesis would be that this population was static in mean endocranial volume. With the computer's help, we can draw random variates from a normal distribution with the assumed standard deviation, assigning them randomly to the times observed for the real fossil sample. Then, we can perform whatever statistic we prefer upon the simulated sample, repeating the process some arbitrarily large number of times. The number of times that meet or exceed the trend observed in the fossil sample provide a p value for the null hypothesis.

What would the result of such a test be for the A. boisei sample? Good question. I'll tell you when I find out.

Why is this important?

The question is really whether the brain size increase in Homo was unique among the early hominids, or whether it may have been replicated in other species. In particular, if the brain size increase also happened in A. boisei in parallel with early Homo, that would be surprising. After all, A. boisei likely had a very different paleoecology than any member of Homo, one that was almost certainly less dependent on technology, less reliant on high-energy foods such as meat, and presenting less of a necessity for group coordination of activities. If brain size increase could occur in a significant way in A. boisei, it really raises questions about the pattern of selection on brain size in hominids.

What could explain an increase in A. boisei? One hypothesis would be energetics. The brain is a great energetic drain, because nervous tissue is very costly. For this reason, there is normally fairly strong selection in favor of smaller brains--because they are more energetically efficient. This selection for smaller brains is opposed by selection for brain functions of one kind or another, because a brain that is too small risks losing some function important for survival or reproduction.

A. boisei clearly differed from earlier hominids in its dietary adaptation, and diet determines the overall energy budget available for an organism. Suppose that the robust masticatory adaptation of A. boisei allowed the species to have a more dependable source of foods during periods of scarcity–because the range of fallback foods was extended into foods unavailable to other hominids, for example. If this were the case, then A. boisei may have had significantly less resource stress during periods of resource scarcity for other hominids, and may therefore have had less trouble meeting their energetic demands. This would mean that the selection against larger brains on the basis of their energetic disadvantages might well be weaker in a robust australopithecine. With other sources of selection on brain function the same–or even possibly increased due to a small reliance on rudimentary toolmaking or other mental adaptations–the brain would increase in size. </p>

Some have used the apparent increase in brain size in A. boisei as an argument to address the importance of brain size expansion in later Homo. This is a point worth addressing, because it is a potentially misleading comparison. One way that it misleads is in the magnitude of change necessary to explain the apparent trends. In A. boisei, a straight regression through the earliest and latest observations indicates an increase in brain size of roughly 70 mL per million years. Of course, this regression like all others is most influenced by the smallest and largest values on the independent axis. Considering the probability that KGA 10-525 was actually larger than its instantaneous average, and that Omo L338y-6 was actually small, the actual amount of change in the species over time was likely much less than 70 mL per million years. A consideration of the data points excluding these extreme values yields a nonsignificant increase of only 21.5 mL per million years.

In contrast, the magnitude of the increase in endocranial volume in Middle Pleistocene humans is much larger. Over the past million years, humans have increased from an average of around 900 mL to the present average of around 1350 mL, for a rate of 450 mL per million years. This is at least fivefold and more probably twentifold higher than the rate in A. boisei, and does not consider the observation that the change was concentrated in the more recent Middle and Late Pleistocene. Moreover, this rate is indeed a difference between early and late average values rather than a regression including early and late extreme values. One might object that we should consider the rate of change relative to the current absolute size rather than the absolute change. From the perspective of selection and the function of brain tissue, this question is not easy to answer: it could go either way. But a strict consideration of relative brain increase as opposed to absolute brain increase still shows that recent humans increased at a rate probably seven to tenfold higher than in A. boisei. And the increase within the past 250,000 years--from approximately 1100 to 1350 mL--would indicate a much higher rate of change, at 1000 mL per million years.

So the observation of a slight trend toward higher brain size in A. boisei would not diminish the impressive degree of change in recent human evolution. Nor does it really lend to the idea that brain increases were widespread among fossil hominids and therefore unsurprising. In all likelihood there were other surprising changes, such as the increase from Australopithecus to Homo, and the increase from H. habilis to early humans. Each of these changes deserves a unique explanation, since the brain is not a character likely to increase in size at random or under the influence of genetic drift. And since the most recent increase in Pleistocene hominids occurred in every inhabited region of the world, it would require either gene flow between regions or several unique cases of simulaneous parallel evolution to explain.

Bottom line: is there anything to explain here in A. boisei? I don't really think so. The apparent trend is too likely to be generated by the outlying observations. Even if a trend existed in the species over time, it appears to have been pretty low in magnitude. This remains a case where the recovery of a single specimen with the right measurements and date would completely eliminate any statistically significant result.


Conroy GC, Weber GW, Seidler H, Tobias PV. 1999. Endocranial capacity of early hominids. Science 283:9.

Elton S, Bishop LC, Wood B. 2001. Comparative context of Plio-Pleistocene hominin brain evolution. J Hum Evol 41:1--27.

Hawks J, Wolpoff MH. 1999. Endocranial capacity of early hominids. Science 283:9b.

Holloway RL. 1980. Within-species brain-body weight variability: A reexamination of the Danish data and other primate species. Am J Phys Anthropol 53:109--121.

Holloway RL, Yuan MS. 2004. Endocranial morphology of A. L. 444-2. In: Kimbel WH, Rak Y, Johanson DC, editors, The skull of Australopithecus afarensis. Oxford, UK: Oxford University Press. p 123--135.

Lee SH, Wolpoff MH. 2003. The pattern of evolution in Pleistocene human brain size. Paleobiology 29:186--196.

Pakkenberg H, Voigt J. 1964. Brain weight of the Danes: forensic material. Acta Anatomica 56:297--307.

Tobias PV. 1971. The brain in hominid evolution. Columbia: New York.