Brain size growth in wild and captive chimpanzees

I’m back in Astana, overcoming jet lag, after the annual conference of the American Association of Physical Anthropologists, which was held in my home state of Missouri. I’d forgotten how popular ranch dressing and shredded cheese is out there; but hey, at least you can drink the tap water! It was also nice to be immersed in a culture of evolution, primates and fossils, something so far lacking at the nascent NU.

Although I usually present in evolution and fossil-focused sessions, my recent interest in brain growth landed me in a session devoted to Primate Life History this year. The publication of endocranial volumes (ECVs) from wild chimpanzees of known age from Taï Forest (Neubauer et al., 2012) led me to ask whether this cross-sectional sample displays the same pattern of size change as seen in captive chimpanzee brain masses (Herndon et al., 1999). These are unique datasets because precise ages are known for each individual, and this information is generally lacking for most skeletal populations. We therefore have a unique opportunity to estimate patterns and rates of growth, and to compare different populations. Here are the data up to age 25 (the oldest known age of the wild chimps):

fig2 raw data copy

Brain size plotted against age in chimpanzees. Blue Ys are the Yerkes (captive) apes and green Ts are the Taï (wild) chimps. Note that Yerkes data are brain masses while the Taï data are endocranial volumes (ECVs). Mass and volume – as different as apples and oranges, or as oranges and tangerines? Note the relatively high “Y” at 1.25 years, who was omitted from the subsequent analysis.

This is an interesting comparison for a few reasons. First, to the best of my knowledge brain size growth hasn’t been compared between chimp populations (although it has been compared between chimps and bonobos: Durrleman et al., 2012). Second, many studies have found differences in tooth eruption, maturation and skeletal growth and development between wild and captive animals, but again I don’t think this has been examined for brain growth. Finally, and most fundamentally, it’s not clear whether ECV and brain mass follow the same basic pattern of change (brain mass but not ECV is known to decrease at older ages in humans and chimps, but at younger ages…?.

So to first make the datasets comparable, I used published data to examine the relationship between brain mass and ECV in primates, to estimate the likely ECV of the Yerkes brain masses. Two datasets examine adult brain size across different primate species (red and blue in the plot below), and one looks at brain mass and ECV of individuals for a combined sample of gorillas (McFarlin et al., 2013) and seals (Eisert et al., 2013). In short, ECV and brain mass in these datasets give regression slopes not significantly different from 1. One dataset has a negative y-intercept significantly different from 0, meaning that ECV should actually be slightly less than brain mass, but I think this pattern is driven by the really small-brained animals like New World Monkeys).


The relationship between endocranial volume and brain mass in primates (and Weddell seals). Solid lines and shaded confidence intervals are given for each regression, and the dashed line represents isometry, or a 1:1 relationship (ECV=brain mass). The rug at the bottom shows the range of the Yerkes masses. Note that the red and black regressions are not significantly different from isometry, while the blue regression is shifted slightly below isometry.

So let’s assume for now that the ECVs of the Yerkes apes are the same as their masses, meaning the two datasets are directly comparable. There are lots of ways to mathematically model growth, and as George Box famously quipped, “All models are wrong, but some are useful.” Here, I wanted to use something that explained the greatest amount of ontogenetic variation in ECV while also levelling off once adult brain size was reached (by 5 years based on visual inspection of the first plot above). This led me to the B-spline. With some tinkering I found that having two knots, one between each 0.1-2.5 and 2.6-5, provided models that fit the data pretty well, and I resampled knot combinations to find the best fit for each dataset. The result:

B-splines describing the relationship between ECV (or brain mass) and age in the TaÏ (green) and Yerkes (blue) data. Although resampling identified different knots for each sample, the regression coefficients are not significantly different.

B-splines describing the relationship between ECV (or brain mass) and age in the TaÏ (green) and Yerkes (blue) data. Note that although the Yerkes line is elevated above the Taï line after 4 years, the confidence intervals (shaded regions) overlap at all ages.

These models fit the data pretty well (r-squared >0.90), and nicely capture the major changes in growth rates. Resampling knot positions reveals best-fit models with different knots for each sample, but otherwise the two models cannot be statistically distinguished from one another: the 95% confidence intervals of both the model coefficients and brain size estimates overlap. So statistical modelling of brain growth in these samples suggests they’re the same, but there are some hints of difference.

Growth rates at each age calculated from the B-spline regressions. Note these are arithmetic velocities and not first derivatives of the growth curves.

Growth rates at each age calculated from the B-spline regressions. Note these are arithmetic velocities and not first derivatives of the growth curves. The dashed horizontal line at 0 indicates the end of brain size growth.

Converting the growth curves to arithmetic velocities we see what accounts for the subtle differences between samples. The velocity plot hints that, in these cross-sectional data, brain size increases rapidly after birth but growth slows down and ends sooner in Taï than among the Yerkes apes. I’m cautious about over-interpreting this difference, since there is great overlap between growth curves, and there is only one Taï newborn compared to about 20 in Yerkes: even just a few more newborns from Taï might reveal greater similarity with Yerkes.

So there you have it, it looks like the wild Taï and captive Yerkes chimps follow basically the same pattern of brain growth, despite living in different environments. Whereas the generally greater stressors in the wild often lead to different patterns of skeletal and dental development in wild vs. captive settings, brain growth appears pretty robust to these environmental differences. That brain growth should be canalized is not too surprising, given the importance of having a well-developed brain for survival and reproduction. But it’s cool to see this theoretical expectation borne out with empirical observations.

2015 AAPA conference: More brain growth

The American Association of Physical Anthropologists is holding its annual meeting next year in St. Louis, in my home state of Missouri (I’m from Kansas City, which is by far the best city in the state, if not the entirety of the Midwest). I’ll be giving a talk comparing brain size growth in captive and wild chimpanzees, on Saturday 28 March in the Primate Life History session. Here’s a sneak peak:

Velocity curve for brain size from birth to 5 years in wild (green) and caprive (blue) chimpanzees. For the captive models, the dashed line is fit to the raw brain masses, and the solid line is fit to the estimated endocranial volumes.

Velocity curves for brain size growth from birth to 5 years in wild (green) and captive (blue) chimpanzees. The wild data are endocranial volumes, but the captive specimens are represented by brain masses. So the captive data are modeled for both the original masses (dashed) and estimated volumes (solid). Wild data are from Neubauer et al. 2011, captive data from Herndon et al., 1999.

Abstract: This study compares postnatal brain size change in two important chimpanzee samples: brain masses of captive apes at the Yerkes National Primate Research Center, and endocranial volumes (ECVs) of wild-collected individuals from the Taï Forest. Importantly, age at death is known for every individual, so these cross-sectional samples allow inferences of patterns and rates of brain growth in these populations. Previous studies have revealed differences in growth and health between wild and captive animals, but such habitat effects have yet to be investigated for brain growth. It has also been hypothesized that brain mass and endocranial volume follow different growth curves. To address these issues, I compare the Yerkes brain mass data (n=70) with the Taï ECVs (n=30), modeling both size and velocity change over time with polynomial regression. Yerkes masses overlap with Taï volumes at all ages, though values for the former tend to be slightly elevated over the latter. Velocity curves indicate that growth decelerates more rapidly for mass than ECV. Both velocity curves come to encompass zero between three and four years of age, with Yerkes mass slightly preceding Taï ECV. Thus, Yerkes brain masses and Taï ECVs show a very similar pattern of size change, but there are minor differences indicating at least a small effect of differences in habitat, unit of measurement, or a combination of both. The overall similarity between datasets, however, points to the canalization of brain growth in Pan troglodytes.

Cryptic Variation and Ice Man

This fortnight’s Current Biology has some interesting articles, two of which caught my attention. First is a “Quick Guide” to cryptic variation, which is genetic variation that goes unnoticed under most circumstances. Also published is the mtDNA sequence of the 5,000 year old Tyrolean Ice Man (a.k.a. Ötzi; this paper actually came out right before Halloween, so I suppose I was too excited about the holiday to write about it then).

First, “cryptic variation.” The authors describe cryptic variation as, “unexpressed, bottled-up genetic potential. … expressed under abnormal conditions such as in a new environment or a different genetic background” (Gibson and Reed 2008). Sounds impossible, because one quickly asks, how can we study ‘cryptic variation’ if it refers to something that is phenotypically unexpressed? But it has been documented in plants and Drosophila, the work-horse-fly of biology. As an example, the authors cite the condition of Antennapedia in Drosophila, in which a mutation causing legs to grow in place of flies’ antennae. When placed into the genomes of different species of Drosophila, this mutation produces different phenotypes. This indicates that variation can be affected by interactions among genes, a phenomenon known as epistasis.

A related phenomenon is ‘canalization,’ which is the evolution of phenotypic ‘buffering’ that prevents variation from arising during development. [For a good synthesis of the concept of canalization, evidence for it, and an application in anthropology, check out Hallgrimsson et al.’s Yearbook paper (Hallgrímsson et al. 2002)] Basically, it seems that enough stabilizing selection can ensure that an individual’s phenotype will develop to a given form in spite of various environmental or internal stresses (i.e. climate and the external environment, or the genetic environment of an organism). This suppression of phenotypic variation can allow ‘cryptic’ genetic variation to accumulate, to be suddenly expressed in a future generation because of certain circumstances. The authors point out that this is possibly problematic because this is not how genes are supposed to work, as far as we know. I think this is an interesting, and potentially very important, avenue of paleoanthropological research, specifically regarding the possibility of hybridization. I’ve written elsewhere, as have others, about the possibility and implications of hybridization on human evolution. Could hybridizing hominin lineages have ‘released’ some type of cryptic variation? An interesting idea, but as always it’s fairly pointless unless it can be tested. And at the moment I cannot think of a way, but I’ll work on it…

In the mean time, researchers have sequenced the mtDNA of the Tyrolean Ice Man. This poor chap, unfortunately for him but fortunately for science, died and ended up in a glacier between Italy and Austria that preserved his soft-tissue very well, some 5000 years ago. What did the study find? Turns out Ice Man’s mtDNA is part of haplogroup K, but has two specific mutations that make his unlike any living mtDNA haplogroup. I seem to remember reading recently about another ancient mtDNA sequence that is unlike anything modern known in modern humans… Oh yes, the 38 ky old Neandertal from Vindija (Green et al. 2008)! If a 5,000 year-old Italian could have belonged to an extinct mtDNA lineage, what does this mean for a similarly ‘extinct’ 38ky old Neandertal? Not a whole lot, but it does underscore how easily mitochondrial lineages can be lost, and it cautions against using the single Neandertal’s mtDNA to argue against their contribution to the modern Homo sapiens gene pool.

Additionally it highlights some of the limitations of genetic studies. Genetic studies like this are limited to the current database of sequences. Ötzi was compared to a sample of some 2000 individuals’ genomes. But there’s always a chance that Ötzi’s ‘extinct’ mitochondrial haplogroup is present but has not yet been sampled. This reminds me of a recent Q&A in Nature entitled, “The pitfalls of tracing your ancestry.” Here, Charmaine Royal of Duke University described issues that arise when people try to trace their ancestry with genetic testing. Here’s what Royal said that has bearing on Otzi, and other ancient genomes:

“The general limitation, I’d say, of all of these tests, is that they can’t pinpoint with 100% accuracy who your ancestors may or may not be. Some people are concerned that the biogeographical ancestry test reifies the notion of race. This is the notion that there are four or five parental groups from which we all came and there are discrete boundaries between these groups. But our genetic research has shown that those boundaries don’t exist.

In lineage testing, where someone is wanting to know which tribe or region in Africa they came from, the information that’s given is based on the present day populations. The names of those groups and those locations have changed over time and so people getting that information about present day Africans and extrapolating to who their pre-middle-passage ancestors may have been — that may not necessarily be accurate. So, those limitations need to be clarified.

Another limitation is that the outcomes of ancestry tests are very much dependent on what is already in a database — who a client’s DNA can be matched to. If a database is not comprehensive some potential matches will be missing, and nobody has a complete database. That’s a major limitation, probably one of the biggest.”

Royal also discusses some interesting issues of when genome testing goes wrong—that is, when people’s genetic results about their identity don’t conform to what they’d expected, how they identify themselves. The piece does a good job illustrating the complex nature of cultural identity and genetic affinity. In the same vein, paternity testing creates the same issues: how one’s social identity/reality can be ripped asunder by a genetic test. So, while genetics and genomics are incredibly valuable scientific avenues, it’s always fun to point out their limitations and adverse effects. Anyway, paleogenomics and cryptic variation are interesting topics that will hopefully continue to be developed and incorporated into Anthropology in the coming years.

Gibson G, and Reed LK. 2008. Cryptic genetic variation. Current Biology 18(21):R989-R990.
Green RE, Malaspinas A-S, Krause J, Briggs AW, Johnson PLF, Uhler C, Meyer M, Good JM, Maricic T, Stenzel U and others. 2008. A Complete Neandertal Mitochondrial Genome Sequence Determined by High-Throughput Sequencing. Cell 134(3):416-426.
Hallgrímsson B, Willmore K, and Hall BK. 2002. Canalization, developmental stability, and morphological integration in primate limbs. American Journal of Physical Anthropology 119(S35):131-158.