For decades, scientists argued about the Moon's ancient magnetic field. Some Apollo samples showed magnetization stronger than Earth's — evidence of a powerful, sustained lunar dynamo. Other samples showed weak magnetization — evidence of a feeble field or none at all. Both camps had real data and reasonable models. The debate persisted because neither side could explain the other's samples.
A team at Oxford, reanalyzing the Apollo basalts' chemistry alongside their magnetization (Nature Geoscience, 2026), found a correlation that resolved both camps simultaneously. Samples with more than 6 weight percent titanium showed strong magnetization. Samples below that threshold showed weak fields. The Moon didn't have a strong field or a weak field. It had a weak field almost all the time, punctuated by bursts of intense magnetism lasting a few thousand years at most — possibly as brief as decades — triggered when titanium-rich material deep inside the Moon melted and temporarily amplified the core dynamo.
The rare bursts are not the surprise. Core-mantle interactions can plausibly produce intermittent amplification. The surprise is why the debate lasted so long. Apollo astronauts selected landing sites on mare basalts — the dark, smooth volcanic plains visible from Earth. Smooth terrain is good for landing. Mare basalts are titanium-rich. Titanium-rich rocks preferentially record the rare magnetic bursts. The sampling was systematically biased toward the anomaly, not the baseline.
The chain of hidden correlation is what makes this instructive. Nobody chose landing sites based on titanium content. Nobody selected samples for their magnetic recording properties. The criterion was terrain smoothness — a logistical decision that seemed orthogonal to any scientific question about magnetic fields. But smoothness correlated with basalt composition, which correlated with titanium enrichment, which correlated with magnetic recording fidelity, which created a sample biased toward events that represent a few thousand years out of billions. The scientific debate was real, the data were accurate, and the fundamental source of confusion was a flight safety decision made before any sample was collected.
The general principle: sampling criteria that appear unrelated to a scientific question can correlate with the answer through hidden intermediate variables. When every step in the chain is reasonable and each correlation is modest, the cumulative bias is invisible to the researchers working with the samples. The remedy is the same as for other measurement traps — diversify the sample. But you can't diversify the sample when you can't return to the source, which is why lunar science waited five decades for the right statistical re-analysis instead of the right new sample.