Can the Most Responsible Scientific Conclusion Be That the Evidence Is Still Uncertain?

Is “We Still Don't Know” Ever a Satisfactory Research Conclusion?

Research is expected to produce answers. Studies are conducted, data are analyzed, papers are written, and readers understandably want to know what the evidence ultimately says.

That expectation can make uncertainty feel like failure. After months or years of research, surely the conclusion should be that an intervention works or does not work, an association exists or does not exist, or one explanation is better than another.

But sometimes the available evidence cannot support any of those statements with sufficient confidence. In such cases, the most scientifically responsible conclusion may be that an important uncertainty remains.

Yes, Uncertainty Can Be the Most Accurate Conclusion

Scientific Responsibility Includes Knowing When the Evidence Cannot Carry a Strong Conclusion

Research conclusions should be calibrated to evidence. When evidence is strong, researchers can make correspondingly confident statements. When important limitations remain, the language and scope of the conclusion should reflect them.

This principle is fundamental to the role of uncertainty in scientific knowledge. Uncertainty is not something added reluctantly after the “real” conclusion. It is part of what the evidence tells researchers.

Uncertainty does not mean complete ignorance

There is a considerable difference between having no evidence and having evidence that remains uncertain.

Suppose several studies suggest that an intervention probably helps, but their estimates are imprecise and some important biases remain. Researchers may reasonably judge that benefit is more plausible than harm while still lacking enough confidence to make a strong claim about the magnitude or reliability of that benefit.

Alternatively, evidence may exclude very large effects while remaining compatible with small benefit, little effect, or small harm. Researchers have learned something, even though several practically important possibilities remain open.

“No evidence of an effect” is not automatically “evidence of no effect”

This distinction is particularly important when studies produce non-significant results.

Imagine a small study estimating a beneficial effect but with a wide confidence interval that is compatible with meaningful benefit, little difference, and some harm. A non-significant statistical test does not establish that the intervention has no effect. The study may simply be too imprecise to distinguish among those possibilities.

Concluding “there is no effect” would therefore convert uncertainty into certainty without evidential justification.

Evidence supporting little or no meaningful effect requires information capable of excluding effects large enough to matter, not merely failure to reject a null hypothesis.

Uncertainty can arise for several different reasons

Saying “the evidence is uncertain” is most useful when researchers explain why.

GRADE provides one structured example. Its assessment of certainty considers risk of bias, inconsistency, indirectness, imprecision, and publication bias. Concerns in one or more of these domains can reduce confidence in an estimate or conclusion.

Those sources of uncertainty have different implications. If evidence is imprecise, more informative data may help. If studies are inconsistent, researchers may need to understand heterogeneity. If evidence is indirect, research more closely aligned with the target question may be necessary. If bias is the central problem, simply increasing sample size may accomplish little.

Conflicting evidence may make uncertainty the correct conclusion

If credible studies produce substantially different results and researchers cannot yet explain why, the evidence may not support choosing one pattern confidently.

This is precisely when conflicting evidence should increase uncertainty rather than force a choice. A conclusion that preserves disagreement can be more accurate than an average or majority judgment that hides it.

Sometimes later research will show that the studies differed because effects genuinely vary among populations or contexts. In that case, what initially looked like uncertainty about one universal answer may eventually become knowledge about several conditional answers.

Low-certainty evidence is not the same as a negative finding

Formal certainty assessments make this distinction explicit. GRADE describes four levels of certainty: high, moderate, low, and very low. These levels concern confidence in the evidence, not whether an effect is positive or negative.

A large estimated benefit can be supported by low-certainty evidence. A near-zero effect can be supported by high-certainty evidence. Direction and certainty answer different questions.

This matters because readers sometimes interpret “low certainty” as though researchers had found no effect. What it actually means is that confidence in the estimated effect is limited and future or better evidence may meaningfully alter the conclusion.

Uncertainty should be specific rather than vague

“More research is needed” is often technically defensible but scientifically uninformative.

A stronger conclusion identifies what remains uncertain. For example:

• The direction of the effect is uncertain because credible studies point in different directions.
• A benefit is plausible, but its magnitude remains uncertain because estimates are imprecise.
• The association is consistently observed, but its causal interpretation remains uncertain because residual confounding cannot be excluded.
• The effect is reasonably established in one population, but applicability to another population remains uncertain.

Each statement tells the reader where the uncertainty sits. That information is considerably more useful than treating uncertainty as one undifferentiated category.

Researchers can be confident about some parts of a conclusion and uncertain about others

Scientific conclusions often contain several layers.

Researchers may be highly confident that an association exists but uncertain about its mechanism. They may be confident that an intervention produces some short-term effect but uncertain about long-term outcomes. They may know the direction of an effect reasonably well while remaining uncertain about its magnitude.

This is why researchers should calibrate confidence to the particular conclusion rather than attach one global label to an entire study or topic.

High uncertainty does not automatically mean no decision can be made

Scientific conclusions and practical decisions are related but distinct. Decision-makers sometimes must act before research uncertainty disappears.

A clinician, policymaker, educator, organization, or individual may need to choose among options despite limited evidence. Such decisions can also depend on potential benefits and harms, costs, feasibility, values, preferences, reversibility, and the consequences of delaying action.

Uncertain evidence should therefore be communicated honestly rather than transformed into a recommendation merely because a decision is required.

The scientific statement may be “we remain uncertain about the effect.” A practical decision can still follow, but it should incorporate that uncertainty rather than pretending it has been resolved.

Uncertainty is not necessarily permanent

An uncertain conclusion identifies what future research could potentially change.

If uncertainty comes from small samples, larger or more informative studies may improve precision. If results vary by context, targeted studies may investigate moderators. If most evidence shares a methodological weakness, a different design may be particularly informative.

This is one reason new evidence can legitimately change previous scientific beliefs. Uncertainty tells researchers where the current evidential structure remains vulnerable to revision.

Uncertainty should not be manufactured merely because science is never absolutely certain

There is an important opposite error. Because empirical knowledge is generally open to revision, someone might claim that every conclusion is equally uncertain.

That is not justified. Bodies of evidence differ greatly in credibility. The possibility that future evidence could revise a conclusion does not mean current evidence provides no reason for confidence.

The National Academies emphasizes that confidence in scientific knowledge can grow through multiple channels of evidence across varied studies. Scientific fallibility and scientific confidence are therefore compatible.

The wording of the conclusion should communicate the level of confidence

Language matters because readers often treat conclusions as stronger than the underlying evidence warrants.

Cochrane guidance links wording to certainty assessments. For example, appropriately qualified language can distinguish findings supported with moderate certainty from those supported with low or very low certainty.

The exact vocabulary will vary by discipline, but the principle is broader: claims should communicate uncertainty rather than bury it in a limitations paragraph after presenting an absolute conclusion.

Compare these statements:

• “The intervention does not improve learning.”
• “The available evidence does not show a clear improvement in learning.”
• “The effect on learning remains uncertain because the available studies are small and their estimates are compatible with both meaningful benefit and little effect.”

Those statements are not stylistic variants. They make different scientific claims.

An uncertain conclusion can be highly informative

A carefully characterized uncertainty can tell researchers which possibilities remain plausible, which have become unlikely, why current evidence cannot distinguish among the remainder, and what type of evidence would be most informative next.

That is genuine knowledge. The conclusion may be less satisfying than a clean yes or no, but the purpose of research is not to maximize decisiveness. It is to make conclusions proportionate to evidence.

When “No Significant Difference” Does Not Mean “No Effect”

The uncertain conclusion is not weaker scholarship. It is a more accurate representation of what the data can and cannot distinguish.

Why Uncertainty Should Not Be Confused With Failure or No Effect

State Exactly What You Are Uncertain About

If your evidence does not justify a definitive answer, do not manufacture one to make the conclusion sound stronger. Instead, identify the remaining uncertainty as precisely as possible.

Ask whether the uncertainty concerns the existence of an effect, its direction, magnitude, causal interpretation, mechanism, generalizability, long-term consequences, or applicability to a particular context.

Then identify why the uncertainty remains. That diagnosis often makes the conclusion more useful because it indicates what kind of evidence could reduce it.

A useful uncertain conclusion is therefore not simply “more research is needed.” It explains what additional evidence needs to resolve and why resolving it would matter.

Before Concluding That the Evidence Is Still Uncertain

Questions About Uncertain and Inconclusive Research Evidence

“The Evidence Is Still Uncertain” Can Be a Scientifically Strong Conclusion

How to Cite This Guide

This guide is intended to be read, shared, and used in research, teaching, and academic work. If you draw on its ideas, explanations, or other content, please acknowledge the source by citing the guide. Doing so gives appropriate credit and helps your readers locate the original resource.