How Scientists Build Cases When Nothing Can Be "Proven"
Your morning yogurt boasts it "supports digestive health." Your cereal claims it "lowers cholesterol." But behind these simple statements lies a high-stakes scientific detective story where absolute proof is a phantom and the rules of evidence are constantly rewritten.
We live in world awash in scientific claims, from food labels promising health benefits to news reports declaring the latest cancer risk. Yet few realize that science never truly "proves" anything in the absolute sense 4 . As Naomi Oreskes, historian of science at Harvard University notes: "Proof—at least in an absolute sense—is a theoretical ideal, available in geometry class but not in real life" 4 . This creates a fundamental tension for scientists compiling evidence dossiers—comprehensive collections of data submitted to regulators to substantiate claims. How do we navigate the gray zone where evidence must be weighed, not counted, and where the goalposts for "sufficient proof" keep shifting?
Modern evidence standards trace back to legal frameworks. The landmark Frye v. United States case (1923) established that expert testimony must be based on principles "sufficiently established to have gained general acceptance" . This case involved a Harvard psychologist named William Moulton Marston (later creator of Wonder Woman), whose early lie detection test was excluded from court. Though Marston's name vanished from legal history, his case established that scientific evidence requires community validation, not just individual conviction .
Consider health claims on foods. The European Food Safety Authority (EFSA) demands "conclusive evidence of cause and effect"—a near-impossible standard for complex nutritional science where:
| Evidence Type | Strength | Limitations in Nutrition |
|---|---|---|
| Randomized Controlled Trials (RCTs) | Gold standard for drugs | Short duration; isolated nutrients ≠ whole foods |
| Cohort Studies | Real-world relevance | Confounding factors; expensive |
| Mechanistic Studies | Shows biological plausibility | May not translate to humans |
| Animal Studies | Controlled conditions | Species differences; ethical limits |
Enter the Burden of Proof methodology, a meta-analytic framework developed for the Global Burden of Disease study. This approach quantifies evidence strength by:
| Stars | Harmful Risks | Protective Factors | Interpretation |
|---|---|---|---|
| No association | No association | Weakest evidence | |
| 0-15% risk increase | 0-13% risk reduction | Weak association | |
| >15-50% increase | >13-34% reduction | Moderate | |
| >50-85% increase | >34-46% reduction | Strong | |
| >85% increase | >46% reduction | Very strong |
For example:
In 1922, psychologist William Marston staged a dramatic demonstration for law students. An actor burst into class posing as a messenger, delivered an envelope, and secretly sharpened a knife against his glove. After the actor left, students recorded their observations.
Results were shocking:
| Metric | Result | Implication for Evidence |
|---|---|---|
| Observable facts | 147 | Baseline for accuracy |
| Average facts recalled | 34 | Extreme information loss |
| Critical events missed | Knife handling (100%) | Failure on key details |
| Accuracy under cross-examination | Significantly decreased | Stress degrades recall |
Marston's experiment exposed the myth of the "reliable eyewitness"—whether human or experimental. Modern dossiers avoid this by:
Weights all available data (not just "positive" studies)
Example: PASSCLAIM criteria for food claims 2
Corrects for study design flaws
Example: Burden of Proof's adjustment for recall bias in dietary studies 3
Maps mechanism from molecule to health outcome
Example: FUFOSE project's "function → benefit" pathway 1
Tests claim interpretation
Example: Qualifying language like "may support" instead of "proven" 1
Translates complexity into accessible scores
Example: Burden of Proof's 1-5★ system 3
Preparing evidence dossiers isn't about chasing unattainable "proof." It's about building a preponderance of evidence using:
As the Burden of Proof team concluded after analyzing 180 risk-outcome pairs: only 12% showed "very strong" evidence (★★★★★), while 22% had only weak associations (★★) 3 . This humility—acknowledging how much we don't know—isn't a weakness of science, but its greatest strength. In a world of contested facts, the most ethical dossiers don't claim certainty. They illuminate the spectrum of evidence, empowering us to make informed choices in the gray zone.