In this 1920 classic, Sewall Wright—one of the architects of the modern synthesis in evolutionary biology—tackles the "nature vs. nurture" debate with mathematical rigor. By comparing highly inbred guinea pig lineages with random-bred controls, Wright introduces the method of path coefficients to quantify how much of a trait’s variation is due to genetics, shared environment, or pure developmental stochasticity.
TL;DR
Sewall Wright uses inbred guinea pig lineages to disentangle the effects of heredity and environment on coat patterns. He demonstrates that while genetics sets the stage, nearly 92% of the variation in highly inbred populations arises from "irregularity in development"—essentially internal biological noise.
What Problem Does This Paper Address?
The paper addresses a fundamental question in biology: Why do individuals with the same parents (or even identical genetics) look different? Specifically, Wright investigates the "piebald" (spotted) pattern in guinea pigs.
At the time, it was unclear if the high variability in these patterns was due to complex Mendelian segregation of many "modifier" genes, external environmental factors (like nutrition or temperature), or something else entirely. Wright sought a way to partition this variance into measurable components to determine the "relative importance" of each.
Methods
Wright’s approach is a masterclass in experimental design and statistical innovation:
- Experimental Breeding: He compared a "control" stock (randomly bred to maintain genetic diversity) with 23 inbred families that had been mated brother-to-sister for over 20 generations. This inbreeding effectively "fixed" the genetics, eliminating hereditary variation within those lines.
- Data Transformation: Wright realized that percentage data (0% to 100% white) is "compressed" at the extremes. To fix this, he mapped these percentages onto a normal probability curve, treating the result as a linear scale of "color potentiality."
- Path Coefficients: He introduced a precursor to modern Structural Equation Modeling (SEM). By drawing causal diagrams of how traits pass from parents to germ cells to offspring, he derived equations to calculate the "degree of determination" for heredity ($h^2$), tangible environment ($e^2$), and developmental noise ($d^2$).
Datasets
The study utilizes a massive longitudinal dataset from the Bureau of Animal Industry, spanning from 1906 to 1918. The primary analysis focuses on:
- Exp. B: A random-bred control stock (986 offspring).
- Family 35: A highly inbred lineage (471 offspring).
- Dorsal Pattern Records: Patterns were recorded using rubber-stamp outlines and quantified via tracing cloth divided into equal areas.
Key Insights and Ideas
- The Limits of Genetics: In the random-bred stock, heredity accounted for only about 42% of the variation. In the inbred stock, where genetic variation was nearly zero, phenotypic variation persisted, driven almost entirely by developmental noise.
- Developmental Noise vs. Environment: Wright distinguished between "tangible" environment (factors like food or weather that affect a whole litter) and "intangible" factors. He found that tangible environment had almost zero effect on the piebald pattern. Instead, the variation was due to "irregularity in development"—stochastic events during ontogeny.
- Predictive Power: Wright used his model to predict how much the standard deviation of the trait should drop if all genetic variation were removed. His calculated "expected" standard deviation (0.610) almost perfectly matched the "observed" value in the inbred line (0.603).
Key Results
The paper provides a quantitative breakdown of the factors determining the piebald pattern:
| Factor | Random-Bred (Control) | Inbred (Family 35) |
|---|---|---|
| Heredity ($h^2$) | 42% | 3% |
| Tangible Environment ($e^2$) | 0% | 5% |
| Developmental Noise ($d^2$) | 58% | 92% |
These results are summarized in the paper's primary data table, which tracks means and standard deviations across generations and sexes.
Limitations and Future Work
- Assumption of Additivity: Wright provisionally assumed that dominance was lacking in the "modifier" genes affecting the pattern. He acknowledged that if perfect dominance existed, the correlation between parent and offspring would be lower ($1/3$ instead of $1/2$).
- The "Black Box" of Development: While Wright successfully quantified "developmental noise," he could not identify its biological mechanism. He labeled it "chance," leaving it to future embryologists to determine what specific cellular or chemical irregularities cause these patterns to shift.
- Methodological Expansion: This paper served as a brief introduction to path coefficients; Wright noted that a more comprehensive mathematical treatment of the method was forthcoming.
Summary updated