The comparative method is simple in principle: you cannot run true experiments on history (cannot rerun 1500 CE with different variables), but you can compare historical cases and observe how outcomes differ when conditions differ. If outcome A appears in every case with condition X, and outcome B appears in every case without condition X, you have evidence that X causes A. This logic applies to history as rigorously as to biology or geology. The method requires: (1) multiple independent cases, (2) identification of variables that differ between cases, (3) observation of outcome differences, (4) inference about causation based on covariation.1
Why this matters: history becomes falsifiable science when you can predict outcomes from variables. "Domesticable animals determine technological development" is falsifiable: if you observe a continent with domesticable animals that remains technologically primitive, the claim is false. If you observe that every continent with domesticable animals develops complex technology, the claim gains support. The comparative method transforms history from narrative ("here's what happened") to science ("here's why it happened, and here's how I'd know if I was wrong").
Independent Cases
Cases must be genuinely independent—not variations within the same system but separate systems that developed autonomously. Five major domestication centers (Fertile Crescent, China, Mesoamerica, Andes, Sub-Saharan Africa) developed independently; one is not caused by another. This independence is valuable because if the same outcome appears across independent cases, causation is more plausible. If only one case showed the outcome, coincidence couldn't be excluded.1
Variable Identification
Which variables to compare? The method requires that variables be (1) observable, (2) quantifiable or categorizable, (3) varying between cases, (4) plausibly relevant to outcomes. "Domesticable animals available" is a good variable: it's observable (you can identify which animals are available), categorical (present or absent), varies between continents, and is plausibly relevant to population density. "Divine will" is a poor variable: it's not observable, not quantifiable, doesn't clearly vary between cases.
Outcome Observation
What outcomes matter? In environmental determinism: technology level (measured by presence of writing, metallurgy, organized warfare), state complexity (measured by hierarchy level, population density), disease environment (measured by genetic resistance alleles). Each outcome should be defined clearly enough that you can measure it across cases.1
Covariation Analysis
Does outcome X consistently appear with variable V? When V is present, does X appear? When V is absent, does X fail to appear? If yes consistently, covariation is established. Covariation plus plausible mechanism plus multiple cases suggest causation. The logic is not airtight (correlation could be coincidence or caused by unmeasured variable), but it's stronger than narrative. The more cases showing the same pattern, the stronger the evidence.
Variable: Domesticable large animals available on continent
Cases:
Pattern: High domesticable animal diversity correlates with state formation and technology advance. Low diversity correlates with less advanced states. The pattern holds across independent cases.
Causation inferred: Domesticable animals enable agriculture → agriculture enables surplus → surplus enables specialization → specialization enables states. Mechanism plus pattern across cases suggests causation.1
Variable: Continental axis orientation (east-west vs. north-south)
Cases:
Pattern: East-west oriented continents show faster technology diffusion. North-south oriented continents show slower diffusion. Axis orientation correlates with diffusion speed.
Causation inferred: East-west latitude bands have similar growing conditions, enabling crop diffusion. North-south crossing requires agricultural reinvention at each latitude. Similar latitudes → easier diffusion → faster accumulation of technology.1
Tension 1: How Many Cases Are Enough?
Five domestication centers isn't 100 experiments. Are five cases sufficient to establish causation? The answer depends on how consistent the pattern is. If all five cases show the same outcome and five is the universe of independent domestication centers, that's fairly strong evidence. But if you're making claims about history generally, five cases seem like a small sample. The statistical principle (larger samples give stronger inference) conflicts with the historical reality (there are only five independent domestication centers in human history).
Tension 2: Can You Ever Isolate Variables?
In true experiments, you hold all variables constant except one. In comparative history, multiple variables differ between cases simultaneously. Eurasia doesn't just have domesticable animals; it has east-west axes, favorable climate, and disease environments. Does domestication cause state formation, or do axes cause it, or disease resistance, or something else? The variables are so entangled that isolating causation is difficult.1
Tension 3: Alternative Explanations
For any observed pattern, you can propose alternative explanations. Eurasia developed more advanced technology—was it caused by domesticable animals (Diamond's claim) or by cultural factors (Orientalism claims) or by geographic isolation enabling protected development (some argue)? The comparative method can't definitively rule out alternatives; it can only show which explanation best fits the observed pattern.
Diamond's use of comparative method is its greatest strength and its primary weakness. Strength: comparing continents systematically reveals patterns narrative history misses; it transforms history into testable science. Weakness: five domestication centers is small sample; variables are entangled; alternative explanations remain possible. Diamond addresses the weakness by emphasizing mechanism (showing how domesticable animals cause states) plus pattern (showing the pattern appears across all major domestication centers) plus multiple independent variables (showing that domesticable animals, axes, and disease all separately predict outcomes). This convergent evidence makes his case stronger than any single variable alone would make it.1
Comparative Method in Evolutionary Biology — Evolutionary biologists use comparative method routinely: comparing species across continents, tracking how traits correlate with environments, inferring evolution from comparative data without ever running experiments. This makes evolutionary biology a natural science despite lacking experimental apparatus. The insight: the comparative method is how scientists handle complex systems they can't experiment on. History is like evolutionary biology: you can't create new continents or rerun the past, so you compare natural cases and infer causation from patterns. Both fields achieve scientific rigor through careful case comparison, not through experimental control. This shows that Diamond's methodological approach is standard across natural science, not something unique to history.
Comparative Institutional Analysis — Sociologists compare institutions (legal systems, economic systems, governance structures) across countries to understand how institutions shape outcomes. Why do some countries develop democratic institutions while others develop autocratic ones? Comparing cases shows that institutional differences correlate with historical, geographic, and economic differences. This is the same method Diamond applies: identify variables (geography, resources), observe outcomes (state formation, technology), infer causation from covariation. The insight: the comparative method is how social science becomes scientific despite inability to experiment.
The Sharpest Implication
If the comparative method is valid, then historical claims can be scientific. This means historical speculation about "what made civilizations great" can be tested: you can identify hypothesized causal variables, compare cases, and ask whether the variable actually predicts outcomes. This is radically different from traditional history, which describes events within cases without comparing across cases. The method enables falsification: if you claim domesticable animals cause advanced technology, you're making a falsifiable claim that can be tested by comparing continents. Many historical narratives are not falsifiable (they explain outcomes post-hoc without predicting), making them unfalsifiable and therefore non-scientific. The uncomfortable implication: much of historical scholarship is not scientific because it describes without predicting.
Generative Questions