Evergreen science library

Understand how evidence works.

Fifteen permanent guides for judging research claims, linked from relevant studies throughout the database.

What Is a Randomized Controlled Trial?

Random assignment makes comparison groups more alike at the start, helping researchers isolate the effect of an intervention.

Peer review is quality control by other specialists—not a guarantee that a paper is correct.

A meta-analysis combines numerical results from multiple studies to estimate an overall effect.

A systematic review uses a documented search and selection process to answer a focused question from all eligible evidence.

Two things moving together does not by itself show that changing one will change the other.

Clinical development moves from early safety and dosing to larger tests of benefit, harms and comparative value.

Mice are powerful biological models, but similarity is selective—not complete.

A small p-value measures compatibility with a statistical null model; it does not measure importance or truth.

A control group estimates what would have happened without the tested intervention.

Sample size affects precision and the chance of detecting effects, but a large dataset is not automatically unbiased.

Publication is a sequence of claims, checks and revisions—not a finish line where uncertainty disappears.

Uncertainty is information about the range of plausible answers, not an admission that nothing is known.

Start with the question and methods, then decide whether the conclusion fits the design and results.

A placebo helps separate the specific effect of an intervention from expectations, attention and natural change.

A cohort study follows a defined group to see how exposures relate to later outcomes.