A long-standing narrative in human evolution suggests that our brains reached their maximum size with early modern humans in Europe around 30,000 years ago, or potentially even earlier in Africa at 100,000 years ago. According to this view, human brain size has subsequently declined, leading to theories about neural reorganization or passive reduction due to overall body size changes. However, recent methodological critiques and advances in ancient DNA research challenge this foundational assumption, suggesting that the apparent decline may be a statistical artifact rather than biological reality.
Methodological Problems in Earlier Studies
The 2022 reassessment by Villmoare and Grabowski of a influential 2021 study by DeSilva and colleagues reveals significant methodological issues that have shaped our understanding of brain evolution. Their analysis identifies three critical problems with previous research claiming Holocene brain reduction:
“Populations from around the globe are lumped together, with only 23 crania sampled over what we would argue to be a critical window with regards to their hypothesis, 5–1 ka, and coming from Algeria, England, Mali, China, and Kenya, among other locations. Later modern human samples are focused on Zimbabwe (at 1.06 ka), the Pecos Pueblo sample from the United States (1 ka), and finally, 165 crania (28% of the total sample) are from Australian pre-Neolithic hunter-gatherer populations and dated in DeSilva et al. (2021) to 100 years ago”.
The first issue involves temporal and geographical lumping. Earlier studies combined cranial specimens from vastly different time periods and locations into broad categories. For example, a 30,000-year-old hunter-gatherer from France might be grouped with a 5,000-year-old early farmer from China and a 200-year-old individual from colonial Australia, all classified as "recent" humans. This approach ignores the fact that major transitions like the adoption of agriculture occurred at different times across the globe, potentially obscuring important regional and temporal patterns.
Sampling bias presents another significant challenge. In the DeSilva dataset, over half of the specimens used to define "modern" brain size came from just two recent collections, heavily representing the last few centuries. Earlier periods were represented by much smaller, scattered samples. This imbalance artificially depressed the calculated "modern" average when compared to the limited Paleolithic sample, creating an illusion of decline. As Villmoare & Grabowski note:
“We also observe that the sample of DeSilva et al. (2021) generates a modern human mean of 1,297 cc in the final 100-year category, which is well below other published estimates of contemporary world-wide modern mean human cranial capacity that range from ~1,340 cc up to ~1,460 cc”
The statistical methods used also proved problematic. Change-point analyses claiming a brain size drop around 3,000 years ago assumed balanced, normally distributed data. When Villmoare and Grabowski examined the actual data distribution, they found it violated these assumptions due to the extreme temporal imbalance and skewed residuals. When they reanalyzed the data using appropriate methods and focusing on the best-sampled period of the last 30,000 years, the significant change-point disappeared entirely.