—
## What “Confirmed” Means in Probabilistic Science
One of the consistent features of news coverage of the 2021 study was the use of the word “confirmed” — as in, “DNA confirms Ernie LaPointe is Sitting Bull’s great-grandson.” This language is both accurate and potentially misleading, and it is worth clarifying what scientists mean by confirmation in a probabilistic context.
Scientific confirmation is not the same as absolute proof. In everyday language, “confirmed” suggests that a fact has been established beyond any remaining doubt. In science, particularly in statistics-based fields like genetic genealogy, “confirmed” or “supported” by evidence means that the data are more consistent with the proposed explanation than with alternative explanations, and that the degree of consistency exceeds whatever threshold the researchers established in advance as sufficient to call the result positive.
The probabilistic relatedness method used in the Sitting Bull study is designed to answer a specific question: given the observed patterns of genetic sharing between the hair sample and Ernie LaPointe, and given the patterns of sharing expected among members of the Lakota Sioux population generally, is the observed sharing more consistent with a great-grandparent relationship or with no specific relationship?
The answer the study reported was: consistent with a great-grandparent relationship. The researchers were “delighted to find a match,” in Willerslev’s words — suggesting that the result was not a foregone conclusion, that it was possible the analysis would have found the patterns inconsistent with the claimed relationship.
This probabilistic framing matters because it means the study’s result is not immune to alternative interpretations. A skeptic could argue that the method, applied to degraded DNA with short fragments and a population comparison group drawn from a community with shared ancestry, cannot fully separate the signal of specific recent relatedness from the background of shared ancestry within the Lakota Sioux population. The researchers designed the method to account for population-level sharing, but the methodological details that would allow external evaluation of whether they did so adequately are in the full paper’s methods section — which was not fully accessible for this article.