Designing a Controlled Experiment

Scientific Methodology

Chapter 3 · Designing a Controlled Experiment

A hypothesis stays untested until it's put through a real, structured comparison. This chapter covers the real building blocks of a controlled experiment through one of history's earliest and most famous examples — including its own honest limitations.

The Real, Foundational Case

James Lind, HMS Salisbury, 20 May 1747

Ship's surgeon James Lind selected twelve sailors suffering from scurvy, at similarly early stages of the disease, and divided them into six pairs. Each pair received a different traditional remedy — cider, elixir of vitriol, vinegar, seawater, a purgative mixture, or two oranges and one lemon daily — while everything else about their care stayed the same.

After six days, only the citrus pair showed real, dramatic improvement; the cider pair improved slightly; the other four pairs showed no real improvement at all. It's now considered one of the first controlled clinical trials in the history of medicine.

The Real Components of a Controlled Experiment

Independent Variable

What's deliberately changed between groups — Lind's own six different remedies.

Dependent Variable

What's measured as the outcome — how much each pair's real symptoms improved.

Controlled Variables

What's deliberately kept the same — disease stage, basic diet, and shipboard conditions across all six pairs.

A Real, Honest Limitation

Lind's own trial is genuinely celebrated, but it wasn't perfect: two sailors per group is a very small sample, and Lind didn't randomly assign sailors to their pairs — he simply grouped them. Randomization, covered next, protects against unknown factors (a sailor's own prior health, diet before boarding) unevenly influencing one group over another purely by chance. Even history's most famous early controlled trial had real, genuine room for improvement by later methodological standards.

Why Randomization Matters

Randomly assigning participants to groups — rather than choosing by hand — spreads out real, unknown differences between individuals (age, health, unmeasured habits) evenly across groups by pure chance, rather than letting a researcher's own unconscious choices skew who ends up where. It's a real, structural safeguard against a confounding variable quietly explaining a result that looks like it came from the actual treatment.

The Real, Sobering Postscript

Despite Lind's own clear real result, the Royal Navy didn't formally introduce citrus rations until 1795 — 48 years later. Good experimental design doesn't automatically translate into real-world action; that gap between evidence and adoption is its own genuine, separate problem.

Hands-On Exercises

Exercise 1

Identify the independent variable, dependent variable, and at least two controlled variables in Lind's real trial, in your own words.

📄 View solution

Exercise 2

Explain, in your own words, a real, specific way the lack of randomization could have affected Lind's own results — what unmeasured factor might have differed between the pairs purely by how he happened to group them?

📄 View solution

Exercise 3

Design a simple, real, controlled experiment (any topic you like) explicitly naming its independent variable, dependent variable, and at least two controlled variables.

📄 View solution

Chapter 3 Quick Reference

James Lind's real 1747 scurvy trial compared six treatments across twelve sailors in matched pairs — one of medicine's first controlled clinical trials
Only the citrus pair showed dramatic real improvement
Independent variable = what's changed; dependent variable = what's measured; controlled variables = what's kept the same
Lind's trial genuinely lacked randomization and used a tiny sample — real limitations even in a celebrated case
Randomization spreads unknown confounding factors evenly across groups by chance
The Royal Navy didn't act on Lind's findings for a real 48 years — good evidence doesn't guarantee real-world adoption