Are Mendel's Data Reliable? The Perspective of a Pea Geneticist

Mendel's data exhibit remarkable agreement to the ratios he predicted. In this article, alternative explanations for this close agreement (that inheritance in pea does not conform to the standard statistical model, that data were omitted, that ambiguous data were categorized to better match predicted ratios, and that some data were deliberately falsified) are tested using approaches that are designed to distinguish between these alternatives. The possibility that garden pea (Pisum sativum L.) naturally produces segregation ratios more closely matching Mendelian expectations than predicted by statistical models is rejected. Instead the opposite is found to be the case, making Mendel's results even more remarkable. Considerable evidence is introduced that Mendel omitted some of his experimental results, but this alternative cannot adequately explain the low average deviation from expectations that is characteristic of the segregation data he presented. An underlying bias in Mendel's data favoring the predicted ratio is present, but my analysis could not clearly determine whether the bias was caused by misclassifying ambiguous phenotypes or deliberate falsification of the results. A number of Mendel's statements are argued to be unrealistic in terms of practical pea genetics, suggesting that his text does not represent a strictly accurate description of his experimental methods. Mendel's article is probably best regarded as his attempt to present his model in a simple and convincing format with a minimum of additional details that might obscure his message.