What Null Results Actually Tell Us About Scientific Progress

Science is often told as a story of discoveries — things found, confirmed, and added to the ledger of knowledge. But a large and underappreciated share of what science actually produces is the null result: the experiment that found no effect, no difference, no signal. How we treat those results reveals something important about how science works and where it can go wrong.

What a null result actually means

A null result is not a failed experiment. It is an experiment that returned a definite answer — the answer "no," or more precisely, "no detectable effect under these conditions." When a well-designed, adequately powered study finds no relationship between two variables, that is evidence. It updates the probability that the relationship exists. The more rigorous the test, the stronger the update.

The confusion arises because people conflate "no effect found" with "nothing found out." These are different. The Michelson–Morley experiment of 1887 is the canonical case: designed to detect the luminiferous aether, it found nothing. That null result didn't close a door — it demolished an entire framework and cleared the ground for special relativity. The experiment is now celebrated precisely because it produced a clean negative.

Most null results don't get that treatment. They sit in file drawers.

The file-drawer problem and what it distorts

The publication bias problem is well-documented: journals preferentially publish positive findings, researchers preferentially submit them, and the cumulative effect is a scientific literature that systematically overrepresents effects that exist and underrepresents effects that don't. This is not a matter of fraud. It emerges from individually rational decisions — a null result is harder to publish, so researchers invest less effort in writing it up.

The downstream consequences are significant. When a meta-analysis aggregates published studies on, say, a nutritional supplement or a psychological intervention, it is drawing from a biased sample. The true effect size, estimated from the published literature, will be inflated — sometimes dramatically. What looks like a robust effect across many studies may be partly an artefact of the studies that never appeared.

This is distinct from the replication crisis, though the two are related. The replication crisis concerns whether individual positive results hold up when retested. Publication bias concerns the shape of the evidence base before replication is even attempted. Both problems stem from treating null results as disposable.

How preregistration addresses the problem

One structural response is preregistration: researchers publicly commit to their hypotheses, methods, and analysis plans before collecting data. This makes it harder to reframe a null result as a design flaw after the fact, and it creates a record of studies that were run regardless of outcome. Journals that offer registered reports — where peer review and provisional acceptance happen before data collection — go further, decoupling publication decisions from results entirely.

The evidence that preregistration helps is itself instructive. Studies comparing preregistered and non-preregistered trials in the same fields consistently find that preregistered studies produce smaller effect sizes and more null results. This is not because preregistration makes experiments worse. It is because preregistration removes the thumb from the scale.

Critics raise a fair point: not all research questions are amenable to tight preregistration, especially in exploratory or observational work where hypotheses emerge from data. The response isn't to abandon preregistration but to be precise about which phase of inquiry a study represents — exploratory or confirmatory — and to treat them accordingly.

What null results reveal about scientific honesty

There is a deeper issue here that touches on the epistemic culture of science. A field that cannot absorb null results without embarrassment is a field that has confused the goal of accumulating knowledge with the goal of accumulating positive findings. These are not the same thing. A result that rules out a hypothesis — cleanly, with adequate statistical power — narrows the space of viable theories. That is exactly what science is supposed to do.

Some fields have moved toward what researchers call null result journals or dedicated repositories for negative data. In pharmacology and clinical research, regulatory requirements already demand that trial results be registered and reported regardless of outcome, a policy that has changed the visible landscape of evidence in those areas. The model is worth extending.

The broader lesson is that scientific progress is not just a sequence of positive discoveries. It is a process of eliminating wrong answers, and null results are the primary instrument for doing that. A scientific culture that treats them as second-class evidence will systematically misread its own state of knowledge — accumulating apparent certainties that rest on a foundation with significant gaps hidden beneath it.

Treating a null result seriously is not pessimism. It is rigor.