This paper, published in the Proceedings of the Royal Society in 1897, represents a landmark attempt by Francis Galton to quantify the elusive mechanics of inheritance. Decades before the modern synthesis of genetics, Galton proposed a mathematical framework to determine how much each ancestor contributes to an individual's traits, validating his theory using a unique pedigree dataset of Basset hounds.
TL;DR
Galton proposes a "statistical law of heredity" where parents contribute 1/2, grandparents 1/4, and great-grandparents 1/8 of an offspring's total heritage. By analyzing the coat colors of over 800 Basset hounds, he demonstrates that this geometric series remarkably predicts the distribution of traits across a population.
What Problem Does This Paper Address?
In the late 19th century, the "how" of heredity was a black box. While Charles Darwin had established the importance of variation, the rules governing how traits were passed down remained speculative. Scientists lacked a way to quantify the influence of a distant ancestor versus a parent.
Galton sought a universal law that could account for the "total heritage" of an offspring. He wanted to move beyond qualitative observations to a predictive model that could be used by breeders, clinicians, and evolutionary biologists to understand the persistence of traits over generations.
Methods
Galton’s approach was both theoretical and empirical:
- The Law of Ancestral Heredity: He hypothesized that the total heritage (1.0) is the sum of a geometric series: $1/2 + 1/4 + 1/8 + 1/16...$ where each term represents the collective contribution of a generation.
- Individual Contribution: Breaking this down further, he argued that each individual ancestor in the $n$-th degree contributes $(0.5)^{2n}$ to the offspring. For example, a single grandparent ($n=2$) contributes $(0.5)^4$ or $1/16$ of the total heritage.
- Statistical Validation: To test this, Galton utilized the meticulous breeding records of Sir Everett Millais, who tracked the coat colors of Basset hounds. Galton categorized the hounds into two groups: "Tricolour" (T) and "Non-tricolour" (N).
- Handling the "Unknown Residue": Because the pedigree data only went back three generations, Galton had to account for the missing influence of more remote ancestors. He calculated a correction factor (0.79) based on the probability that a tricolour parent would have tricolour offspring, allowing him to "fill in" the gaps of the infinite series.
Datasets
The study relied on the Basset Hound Club Rules and Stud-Book (1874–1896).
- Total Sample: 817 hounds with known parental colors.
- Sub-samples: 567 hounds with known grandparents and 188 with known great-grandparents.
- Trait focus: The presence or absence of "melanism" (tricolour vs. lemon-and-white).
Key Insights and Ideas
- Meiosis in Math: Galton noted that his law was "strictly consonant" with the binary subdivision of germ cells. The $1/2$ reduction at each ancestral step mathematically mirrors the chromosomal halving we now recognize in meiosis.
- The Power of Averages: Galton explicitly stated that his law relates to average results. While individual "prepotencies" (what we might now call dominant alleles) exist, they are smoothed out across a large population, allowing the statistical law to hold.
- Maternal Prepotency: Interestingly, Galton found that the dam (mother) appeared slightly more influential than the sire (father) in transmitting color, at a ratio of roughly 6:5. However, he chose to ignore this in his final model to maintain the mathematical elegance of the general law.
- Sensitivity of the Model: Galton experimented with slightly altering his coefficients and found that the accuracy of his predictions dropped significantly. This suggested that the $1/2, 1/4, 1/8$ ratios were not just arbitrary, but captured a fundamental truth about the data.
Key Results
The predictive power of the model was startlingly high for the era:
- Grand Totals: In a test of 567 hounds, the model calculated that 391 should be tricolour; the observed number was 387.
- Great-Grandparental Data: In a smaller set of 188 hounds, the model calculated 180 tricolours vs. 181 observed.
- Mating Combinations: The law successfully predicted the outcomes of various parental combinations (e.g., T x T, T x N, N x N) with errors often as low as 2%.
Limitations and Future Work
- Simplified Traits: The model works best for binary traits (T vs. N). Galton acknowledged that more complex, continuous traits (like human stature) require more sophisticated handling of deviations from the mean.
- Neglect of Dominance: The law does not account for what we now know as Mendelian dominance. It treats inheritance as a blending of ancestral contributions rather than the interaction of discrete alleles.
- Data Depth: Galton lamented the difficulty of obtaining deep pedigree data. He suggested that future work should incorporate "collateral data" (siblings and cousins) to further refine the 0.79 correction factor for unknown ancestry.
Note: The figures provided in the metadata for this document (Figures 1 and 2) pertain to separate papers in the same volume regarding the dielectric constants of organic bodies and are not relevant to Galton's study on heredity.
Summary updated