Editor’s Note: This is the first post in Benchling’s new blog series. Join us for stories that inspire discussion and explore the intersection of science, society, and beyond!
While today’s scientists have the advantages of high-throughput sequencing and innovative software, the study of genetics began with a far more humble tool—peas. In 1856, before the concept of genes (or even evolution) existed, an Austrian friar initiated a series of breeding experiments to study how traits are inherited. Eight years and a staggering 29,000 pea plants later,1 Gregor Mendel published his meticulous observations, from which he deduced the basic laws of genetic inheritance. It would take several more decades before anyone realized its importance, but from then on, Mendel’s work has become the foundation for modern biology and genetics.
Blessed with a green thumb and a saintly portion of patience, Gregor Mendel grew—and counted—thousands of garden peas on a 5-acre plot at the St. Thomas Abbey.
Today, Mendel is undisputedly recognized as “the father of genetics.” Still, the quality of his scientific practices has been a long-standing topic of fierce debate. Though Mendelian genetics is ultimately correct, the numbers he reported seem a little too correct; decades of statisticians have argued that he may have falsified and exaggerated his data to match his theory more perfectly. So was Mendel really a fraud?
† See footnotes for a more in-depth explanation.
With the advent of rigorous mathematical statistics, early 20th century statisticians like Ronald Fisher were finally able to apply their tests to Mendel’s data. Indeed, many of the arguments against Mendel cite the statistical tests performed by Fisher. No one doubts the validity of Mendel’s final conclusions, but Fisher suggested that Mendel’s data matched too closely with the theoretical values.
More specifically, Fisher noted that Mendel’s experimental design (namely sample size) would have introduced a significant amount of error into his data; after adjusting for this inherent error, Fisher found that the data matched Mendel’s expected value significantly better than Fisher’s error-corrected value.2 Conspiracy theorists have long since latched onto this particular argument, but as with all mathematical models, it’s only as good as its assumptions. These calculations were based on a specific sample size that was inferred from Mendel’s original paper. In response, Mendel supporters propose that the text is ambiguous, and actual sample size could easily have been larger than suspected—thus negating Fisher’s main contention.3
Much of the Mendel (left) controversy centers around the statistical analyses, conducted decades later, by Ronald Fisher (right). And yes, that’s the same Fisher responsible for Fisher’s Exact Test, Fisher’s Equation, and ANOVA, among others.
Bored with this debate yet? Even with such a mundane explanation for Fisher’s criticism, this is actually one of the least resolved arguments. While there are many more allegations against Mendel, most others have been satisfactorily debunked; in fact, the most exaggerated data in this entire discussion seems to be coming instead, from Mendel’s opponents.
Still, in the last fifty years, more than fifty papers have been published on this “controversy,” 1 including some with particularly inflammatory titles:3
- "Too Many Small χ2’s or Hanky-Panky at the Monastery?" (1968)
- "Great Fakes of Science" (1977)
- "Betrayers of the Truth" (1983)
In actuality, Fisher himself may not have expected the furor that his publication caused; based on the majority of other analyses he conducted, he truly believed in Mendel’s scientific integrity, and even speculated that one of Mendel’s assistants could have tweaked the data.2
Even with reasonable explanations for the supposed discrepancies, the suspicions surrounding Mendel’s reputation have yet to subside—which is perhaps, indicative of a growing sentiment in the scientific community. In the last few decades, instances of scientific misconduct (and retracted publications) have increased dramatically. Indeed, when surveyed, nearly 2% of scientists admitted to fabricating / falsifying / modifying data; another 34% admitted to other “questionable research practices.”4 From the recent scandal of acid-induced stem cells, to the fabricated paper linking vaccines to autism, scientific misconduct is a critical concern with very real, damaging consequences.
The anti-vaccination movement arose from a falsified, now-retracted paper. Even now, without a shred of scientific evidence, the movement continues. Last March, public anti-vaxxer activist, Jenny McCarthy, got slammed on Twitter for her unsupported claims.
At least in the case of Gregor Mendel, his scientific conclusions were correct! Still, if the last 150 years have proven anything, it’s that people love to speculate. And since Mendel’s unpublished records were destroyed by fire, we can keep on guessing.
Too bad Mendel didn't have his data in the cloud.
1. Pires AM, Branco JA. (2010) A statistical model to explain the Mendel-Fisher controversy. Statistical Science 25(4):545-565.
2. Fisher RA. (1936) Has Mendel’s work been rediscovered? Reproduced from Annals of Science v1:115-137.
3. Fairbanks DJ, Rytting B. (2001) Mendelian controversies: a botanical and historical review. Am J Bot 88(5):737-52.
4. Fanelli D. (2009) How many scientists fabricate and falsify research? A systematic review and meta-analysis of survey data. PLoS One 4(5):e5738.
5. Hartl DL, Fairbanks DJ. (2007) Mud sticks: on the alleged falsification of Mendel’s data. Genetics 175(3):975-9.
† More (jargon-filled) details for the aficionados:
The Data: When performing a heterozygous cross (Aa x Aa), the expected phenotypic ratio of dominant to recessive traits is 3:1. Theoretically, the genotypic ratio is 1AA : 2Aa : 1aa, which Mendel attempted to identify by self-pollinating each of the plants with dominant phenotypes. If any of the resulting progeny had the recessive trait, the parent would be classified as heterozygous (Aa). Mendel’s results match this 1AA : 2Aa ratio, nearly perfectly.
The Argument: Fisher argues that there should be a misclassification error, depending on the number of progeny analyzed—if the number of progeny analyzed is too small, the heterozygotes (Aa) may not produce any progeny with the recessive trait (aa) and thus will be misclassified as a homozygote (AA). If n=10 progeny are cultivated from each plant, as described by Mendel, then there’s a 5% chance that all 10 progeny display the dominant phenotype (because 0.7510 = 0.05).2 Accounting for this inherent error, the “AA : Aa” ratio that Mendel should have perceived is instead 1:1.7.
The Rebuttal: This argument depends on the number of progeny examined; as n increases, this misclassification error decreases (e.g. for n=15, error decreases to 1% (0.7515 = 0.01)). Although Mendel indicates that “10 seeds were cultivated” from each plant, he probably harvested and planted more than 10 seeds for each parent, to make sure that he had 10 full-grown progeny that survived. Some of the traits (e.g. seed shape, color) can actually be scored before they grow into “cultivated” seedlings, which means he probably examined more than 10 progeny. This would significantly minimize Fisher’s proposed misclassification error. Therefore, it’s less suspicious that Mendel’s data were closer to the theoretical ratio (1:2) instead of Fisher’s “error-corrected” ratio (1:1.7).
For a full account of all the charges laid against Mendel, see Fairbanks & Rytting (2001)