The disappearance of numerous North American megafauna (animals weighing ≥44 kg) at the end of the Pleistocene (~11,700 years before present [BP]) remains a contentious issue. Three primary hypotheses attempt to explain this extinction event: human overhunting (the 'overkill' hypothesis, with variations like 'blitzkrieg' and 'sitzkrieg'), climate change, or a combination of both. Paul Martin's overkill hypothesis, suggesting overhunting by newly arrived humans, has fueled extensive research and debate since the 1960s. However, arguments against overkill cite the scarcity of megafauna kill sites and inconsistencies between the last appearance datums (LADs) of megafauna and the first appearance datums (FADs) of humans in the Americas. Conversely, evidence supporting climate change as a primary driver includes ancient DNA studies showing genetic diversity loss in some species before human arrival, extinctions among other taxa (birds, reptiles, plants), and body size reductions in surviving megafauna. Radiocarbon dating has played a significant role in testing these hypotheses, with LADs and FADs often used to assess the temporal relationships between human arrival and megafauna extinctions. However, limitations of LAD-based studies include the Signor-Lipps effect (where the last fossil does not represent the last individual) and the uncertainty inherent in radiocarbon dating. To address these limitations, studies have utilized summed probability density functions (SPDFs) as proxies for population fluctuations. However, SPDFs conflate process variation with chronological uncertainty, leading to potentially misleading results. This study employs a more sophisticated method, REC modelling, to overcome these issues and to better understand the relative roles of climate change and human activity in megafaunal extinctions.

Existing research on Late Pleistocene megafauna extinctions is largely divided between proponents of the overkill hypothesis and those emphasizing climate change. The overkill hypothesis posits that human hunting pressure, potentially exacerbated by factors like fire and habitat fragmentation, directly caused megafauna extinctions. However, studies employing summed probability density functions (SPDFs) have yielded mixed results, with some suggesting megafauna population declines predating human arrival in some regions. These studies acknowledged several problems associated with using SPDFs, including radiocarbon sample quality, uncertainties in chronological dating, and potential biases related to taphonomic processes (preservation and loss of samples over time). While recognizing these issues, SPDFs' use in analyzing extinction dynamics remained prevalent, creating the need for a methodology that more explicitly addresses the uncertainties of radiocarbon dating.

This study utilizes a recently developed Bayesian regression technique, Radiocarbon-dated Event Count (REC) modeling, to analyze the largest published database of North American megafauna radiocarbon dates. REC models address the limitations of previous approaches by explicitly accounting for chronological uncertainties inherent in radiocarbon dates. This approach generates a Radiocarbon-dated Event Count Ensemble (RECE), which represents a sample of plausible event count sequences given the uncertainties in the radiocarbon dates. Each sequence in the RECE is then used as the response variable in a separate regression model. This process yields a set of parameter estimates that reflect both sampling variability and chronological uncertainty. The researchers used this method to analyze a database of 521 radiocarbon-dated megafauna samples, controlling for taphonomic bias using an established proxy. Three sets of models were run: 1) megafauna population size compared to human population size (using radiocarbon dates from human sites), 2) megafauna population size compared to the NGRIP δ¹⁸O record (a proxy for climate change), and 3) megafauna population size compared to both human population size and climate change. The analysis was conducted across all megafauna collectively and also broken down by individual taxa (horse, mammoth, mastodon, ground sloth, saber-tooth cat) and regions (Great Lakes, Southwest). A supplementary analysis was also performed using a "chronologically cleaned" dataset to address potential biases associated with multiple dates from a single individual or event, resulting in a database of 432 dates. To further enhance the robustness of their analysis, the researchers extended the climate analysis using the raw, annually resolved NGRIP δ¹⁸O record, incorporating chronological and measurement uncertainty, and extending the time period to 20.0–10.0 ka to examine a longer time frame.

The analysis revealed no significant correlation between human population levels and megafauna population sizes across all models, whether considering all megafauna together or individual taxa and regions. The taphonomic proxy also showed no significant effect, indicating that taphonomic processes did not create a clear temporal trend in megafauna sample frequency. In contrast, all models showed a statistically significant positive correlation between the NGRIP δ¹⁸O record (climate proxy) and megafauna population sizes. This positive correlation implies that decreases in global temperatures were associated with declines in megafauna populations. This finding was consistent across all analyses, including the extended analysis using the annually resolved NGRIP data. The extended analysis, which incorporated the annually-resolved NGRIP δ¹⁸O record and accounted for chronological and measurement uncertainty, supported these findings, suggesting the relationship between megafauna population levels and the NGRIP proxy was robust. The effect size remained consistent across various analyses, with most models having a posterior mean for the climate parameter around 0.1 or higher. The researchers used RECEs to represent the uncertainty range. The positive correlation suggested that decreases in global temperature corresponded with decreases in megafauna population sizes. Final declines in several species coincided with the onset of the Younger Dryas (YD). The researchers provide detailed regional analysis in the Great Lakes and Southwest, showing similar correlation between climate changes and megafauna.

The findings contradict simple overkill models attributing megafauna extinctions solely to human hunting pressure. Although earlier studies using SPDFs suggested a combined effect of human activity and climate change, this study's use of REC modeling, which accounts for chronological uncertainties, reveals a consistent correlation only between climate change and megafauna population levels. The discrepancy between this study's results and earlier findings likely stems from the limitations of SPDFs in accurately representing population dynamics. The study acknowledges the limitations of the radiocarbon record, including sample size limitations and potential biases. Although corrections were made for taphonomic and sampling biases, the authors note the possibility of remaining uncontrolled biases. The results suggest several alternative hypotheses, including the potential for indirect human impacts (e.g., depletion of keystone megaherbivores leading to cascading effects or habitat fragmentation) rather than direct overhunting by a rapidly growing human population. The study highlights the significant role of the Younger Dryas (YD) period, with its abrupt cooling, increased seasonality, increased CO2, and major vegetation changes, in driving megafauna declines.

This study's quantitative analysis using REC modeling, which accounts for chronological uncertainty, provides strong evidence that climate change, specifically the conditions associated with the Younger Dryas, was a primary driver of Late Quaternary North American megafauna extinctions. While the possibility of indirect human impacts cannot be entirely ruled out, the data do not support the hypothesis of widespread overhunting by rapidly expanding human populations. Future research should focus on improving data quality and developing more nuanced models to investigate the complex interplay between human activities and environmental changes in shaping megafauna population dynamics.

The study acknowledges several limitations. The available radiocarbon data, while the largest assembled to date, still suffers from sample size limitations and potential biases related to taphonomic processes and spatial sampling. Although efforts were made to address taphonomic bias, the possibility of remaining uncontrolled biases exists. The analysis primarily focuses on temperature as a climate proxy, neglecting other potentially important climatic factors. Furthermore, the study's focus is on population-level changes, offering limited insight into the specific mechanisms driving extinction within individual species.