03 · What You Need to Know
Why Confirming Existing Findings Can Advance Knowledge
Scientific knowledge depends on cumulative evidence
A published study provides evidence for a claim; publication does not make the claim permanently settled. Empirical results contain uncertainty, and individual studies can be influenced by sampling variation, measurement choices, study design, analytical decisions, context, and other factors.
That is why additional evidence matters. The National Institutes of Health describes replication studies as a core part of the scientific process and states that replication can increase confidence in research findings, experimental methods, and theories. It can also help evaluate validity, generalizability, alternative explanations, and contextual influences.
A confirming study therefore does not merely repeat a sentence already present in the literature. When it is informative and appropriately designed, it changes the body of evidence supporting that sentence.
Confirmation can increase confidence without proving a claim
Suppose one well-designed experiment reports an effect. A second independent study addressing the same question obtains evidence consistent with it. Researchers now have two sources of evidence rather than one.
That can reasonably increase confidence in the finding, particularly if the second study was capable of providing a strong test. But confirmation should not be described as proof. Another study can strengthen evidence without establishing that a claim is universally or permanently true.
NIH explicitly cautions that a successful replication does not necessarily confirm an original result or theoretical finding in an absolute sense; rather, it can provide additional support. Likewise, an unsuccessful replication does not automatically establish that a finding is false.
Confirming evidence
New evidence is consistent with a previous finding and can increase confidence in the claim when the study provides an informative test.
Proof
A much stronger claim implying that the question has been conclusively settled. Empirical confirmation generally does not justify this interpretation.
The value of confirmation depends on how uncertain the claim was beforehand
Not every confirmation adds the same amount of information. Imagine an effect supported by one small exploratory study. A rigorous independent study could substantially alter confidence in that effect.
Now imagine a mature finding supported by many high-quality studies conducted across relevant conditions, with precise and consistent estimates. Another nearly identical study may add comparatively little.
This means the value of confirmation is conditional on the state of the evidence. The question is not merely, “Has this been studied before?” It is, “How much useful uncertainty remains, and can this study reduce it?”
Confirmation can be especially important for influential findings
The consequences of uncertainty are not equal for every research claim. Some findings become foundations for theories, methods, interventions, policies, technologies, or subsequent research programs.
Nature Human Behaviour has argued that scientific progress occurs through confirmation and disconfirmation as well as discovery. In discussing replication studies of influential research, the journal stated that if an original study is highly influential, a rigorous study replicating its methodology can itself be highly valuable.
This suggests a useful prioritization principle: confirmation becomes more consequential when many later claims or decisions depend on the result being reliable.
Confirmation can improve the estimate, not just repeat the direction
Researchers sometimes reduce confirmation to whether a new study obtains a statistically significant result in the same direction as the original. That is too narrow.
A new study can provide another estimate of the magnitude of an effect and its uncertainty. This matters because original estimates can be imprecise or unusually large or small because of sampling variation and selective publication processes.
The Open Science Collaboration's large replication project in psychology illustrates why effect estimates matter. Across 100 replication attempts, replication effects were on average about half the magnitude of the original effects, while different indicators of replication success produced different percentages. The study therefore showed why simply asking whether a result was “significant again” does not capture the entire evidential comparison.
A useful confirmation study should therefore report effect estimates and uncertainty and interpret them alongside the existing evidence rather than treating a matching significance decision as the sole criterion.
Confirmation in another population can test how broadly a finding applies
A finding may be well supported in one population while remaining uncertain elsewhere. Confirmatory evidence from another population can therefore contribute by testing the scope of the result.
For example, Nature Human Behaviour highlighted a large multicultural replication of prospect theory findings across 19 countries as valuable evidence about their reliability beyond the original context. The journal explicitly rejected the idea that a replication becomes less publishable merely because it confirms the previous findings.
But a different population does not automatically make confirmation important. As explained in the guide on testing an existing finding in a new population, the population difference should address a meaningful question about applicability, variation, or generalizability.
Confirmation under different methods can test robustness
Another study can support an underlying claim while changing how that claim is tested. When theoretically appropriate methods produce converging evidence, researchers may become less concerned that the finding depends on one particular measurement, operationalization, analytical choice, or experimental setup.
This is one purpose of conceptual replication. NIH defines conceptual replication as testing the underlying hypothesis of an earlier experiment using a different method or experimental setup.
However, methodological changes also complicate interpretation. If the second study disagrees with the first, researchers must determine whether the underlying claim is weaker than expected or whether the changed implementation tested something importantly different. The choice between direct and conceptual replication should therefore follow the uncertainty you want the new evidence to address.
Independent confirmation can be particularly informative
Evidence produced by an independent research team can address some concerns that repeated findings from the same laboratory might share implementation habits, analytical choices, contextual features, or other dependencies.
Independence does not guarantee correctness, and a study from another team can still share limitations with the original research. But replication across independent teams can provide a stronger test of whether a finding depends on features specific to one research group.
NIH includes control for contextual artifacts, different research personnel, and external validation among the possible functions of replication research.
Confirmation is valuable partly because not every published result replicates
If published findings were automatically reliable, there would be little reason to gather independent confirming evidence. Large replication projects demonstrate why that assumption is unsafe.
The Open Science Collaboration attempted replications of 100 experimental and correlational psychology studies. Although 97% of the original studies reported statistically significant results, 36% of the replications obtained statistically significant results, and replication effect sizes were substantially smaller on average.
More recently, a large investigation reported replications of 274 positive-result claims from 164 papers in the social and behavioural sciences. Of those claims, 55.1% produced statistically significant replication results in the original pattern, despite the replication studies being highly powered on average to detect the original effect sizes.
These results should not be generalized mechanically to every discipline or used to assume that an individual published finding is unreliable. They demonstrate a narrower point: independent confirmation can provide information that the original publication alone does not contain.
Confirmation can be valuable even when the result is unsurprising
Researchers sometimes assume that a confirming result is less valuable because it is less newsworthy. That creates an undesirable asymmetry: contradictory replications appear interesting while successful replications disappear from view.
Nature Human Behaviour has explicitly warned against this tendency. In its editorial on replication, it described a concerning form of publication bias in which replications that contradict original findings may be perceived as more newsworthy than replications that confirm them. The journal argued that rigorous replication contributes to the scientific record regardless of outcome.
The scientific value of a study should therefore not depend on whether the result produces a dramatic reversal.
Confirmation does not automatically mean redundancy
Two studies can reach similar conclusions while differing greatly in how much evidence they provide. A larger sample, stronger measurement, preregistration, better documentation, an independent team, a different relevant context, or a design that addresses a limitation of the original study can make the second result substantially informative.
Conversely, repeatedly conducting weak studies that add almost no precision or independent information can become redundant even if each produces new data.
The distinction is therefore between useful confirmation and unnecessary repetition. Determining when similar research becomes redundant requires looking at the information added, not merely the similarity of the research question.
Watch Out
Do not describe a confirming study as proving that the original finding is correct. Confirmation should update confidence in light of the new evidence, study quality, effect estimates, uncertainty, methodological differences, and the wider literature. Scientific conclusions should rest on cumulative evidence rather than a vote counting how many individual studies were “positive.”
A confirming study can still be original research
Confirmation may involve a familiar question and an unsurprising result, but the evidence itself can be new. A replication that collects new observations is another empirical investigation.
This is why replicating an existing study can constitute original research. Originality does not require the result to contradict previous evidence or the hypothesis to be unprecedented.
Whether that contribution is sufficient for a particular thesis, journal, grant, or other purpose remains context-dependent. An evaluator may have additional expectations regarding scope, originality, significance, or methodology.
