Does Research Discover Facts or Reduce Uncertainty?

Is Research Really a Search for Facts?

Research is often described as a process of discovering facts. Sometimes that description works perfectly well. Researchers can document that an event occurred, establish a measurement, identify a previously unknown organism, determine the sequence of a genome, or record a pattern that had not previously been observed.

Yet many research questions are not resolved by discovering a single fact.

Researchers may instead be estimating how large an effect is, comparing explanations, determining how consistently a relationship occurs, predicting an outcome, interpreting an experience, or establishing the conditions under which a phenomenon changes. In such cases, research often advances knowledge by reducing uncertainty rather than replacing complete ignorance with complete certainty.

Research Can Discover Facts, but It Often Advances Knowledge by Reducing Uncertainty

Research Does More Than Add Facts to a Stockpile

Some Research Questions Really Do Concern Facts

There is nothing inherently wrong with describing some research outputs as factual.

A researcher may establish the date recorded on a historical document, measure the chemical composition of a sample, document the presence of a species at a location, determine whether a particular policy contains a provision, or observe that a specified event occurred under controlled conditions.

Such observations can become extremely well established when the measurement or documentation is reliable and independently verifiable.

The difficulty arises when the language of “discovering facts” is used as a complete description of research. Many questions require more than establishing isolated observations.

Facts and Explanations Operate at Different Levels

Suppose researchers reliably observe that students who attend more classes tend to obtain higher examination scores.

The observed association may be a well-supported empirical finding. But explaining that association raises additional questions. Does attendance improve learning? Are more motivated students both more likely to attend and more likely to study? Do prior achievement, employment, health, or other factors contribute to the relationship?

Accumulating accurate observations is necessary for many forms of inquiry, but the move from observation to explanation requires inference.

This is why moving from observation to explanation is a distinct part of knowledge production rather than an automatic consequence of collecting more facts.

Uncertainty Means More Than “We Do Not Know”

Scientific uncertainty is not a single state. Researchers may know a great deal about a phenomenon while remaining uncertain about particular aspects of it.

They may be uncertain about an exact numerical value, the magnitude of an effect, which of several mechanisms is responsible, how widely a result generalizes, what will happen in the future, or how well a model represents a complex system.

Research can therefore reduce one form of uncertainty while leaving another largely intact.

Research Can Narrow a Range Without Identifying One Perfectly Certain Answer

Imagine researchers trying to estimate the average effect of an intervention. Before adequate research exists, plausible estimates might span a very wide range. A rigorous study may substantially narrow that range while still leaving uncertainty about the exact value.

The knowledge gain is real even though the answer is not perfectly precise.

This logic appears throughout empirical research. Researchers estimate population characteristics from samples, infer mechanisms from observations, predict future outcomes using imperfect models, and compare explanations that may each retain some plausibility.

The National Academies has distinguished uncertainty arising from incomplete knowledge from uncertainty arising through inherent variability. Additional data and improved models can reduce some forms of uncertainty, while other forms cannot be eliminated entirely.

Evidence Can Change the Relative Plausibility of Explanations

Consider two competing explanations for the same phenomenon. Research does not necessarily have to demonstrate that one is absolutely true and the other absolutely impossible to make progress.

A well-designed study might generate evidence that is much more compatible with one explanation than another. Another study may eliminate an alternative explanation. A later investigation may identify a condition under which the favored explanation does not apply.

Knowledge advances because the space of plausible explanations becomes better constrained.

This is closely related to why research generally produces evidence rather than absolute certainty. Evidence allows researchers to update what conclusions deserve confidence.

Uncertainty Can Be Quantified in Some Research

In quantitative research, some uncertainty can be expressed numerically. Estimates may be accompanied by standard errors, confidence intervals, credible intervals, prediction intervals, probability distributions, or other measures appropriate to the analytical framework.

These quantities do not represent every source of uncertainty. A narrow interval around an estimate does not automatically address biased sampling, invalid measurement, incorrect model specification, unmeasured confounding, or limited generalizability.

Quantification is therefore useful but partial. Researchers must also consider forms of uncertainty that cannot be summarized by a single number.

Some Uncertainty Is Epistemic and May Be Reduced With More Knowledge

One useful distinction separates uncertainty caused by incomplete knowledge from uncertainty associated with inherent variability.

Epistemic uncertainty can arise because researchers lack sufficient data, do not fully understand a mechanism, use an imperfect model, or have not adequately measured relevant variables. Additional observations, improved measurements, stronger designs, and better models may reduce this uncertainty.

Research is particularly powerful here because new evidence can constrain possibilities that were previously difficult to distinguish.

Some Variability Cannot Simply Be Researched Away

Other uncertainty reflects variability in the phenomenon itself. Individuals differ. Environments fluctuate. Future events may depend on processes with genuinely variable outcomes.

In such situations, more research can improve our understanding of the distribution of possible outcomes without enabling perfect prediction of each individual outcome.

For example, researchers may estimate with considerable precision how frequently an outcome occurs in a population while remaining unable to predict with certainty which particular person will experience it.

Researchers Can Be Very Certain About Some Things and Less Certain About Others

A mature field may contain claims supported by overwhelming evidence alongside active uncertainty about finer details.

There is no contradiction in saying, for example, that researchers are highly confident a phenomenon occurs while continuing to debate its precise magnitude, mechanisms, boundary conditions, or consequences in particular contexts.

Scientific knowledge should therefore not be classified simply as “known” or “unknown.” Researchers must decide how confident to be in particular conclusions, and those confidence judgments can differ across claims within the same field.

Reducing Uncertainty Can Reveal New Uncertainty

Research sometimes answers one question only to expose another.

A study may establish that an intervention has an effect but reveal unexplained differences among participants. Researchers may identify a mechanism but discover that it behaves differently under particular environmental conditions. Better measurement may resolve an old controversy while revealing variation that previous instruments could not detect.

This is not a failure of knowledge production. Greater resolution can make previously invisible questions researchable.

Research Knowledge Can Therefore Change Without Becoming Arbitrary

If research reduces uncertainty rather than delivering immutable answers, later evidence can change what researchers conclude. New measurements may be more accurate. Larger studies may produce more precise estimates. New populations may reveal boundary conditions. Alternative explanations may become more or less plausible.

That is why scientific knowledge can change when new evidence appears.

Revision does not mean that researchers simply replace one opinion with another. Ideally, conclusions change because the evidential basis for judging them has changed.

Research Can Improve an Answer Without Making It Perfectly Certain

The study has produced knowledge even though it has not identified an eternally fixed answer. Several previously plausible possibilities have become less plausible, and the remaining possibilities are better constrained.

Common Misunderstandings About Facts and Uncertainty

Ask What Your Study Makes Less Uncertain

When thinking about your contribution, you do not always need to ask, “What new fact did I discover?” A more useful question may be, “What uncertainty does this study reduce?”

This framing is particularly helpful when your research estimates effects, compares explanations, examines mechanisms, studies complex social phenomena, or investigates questions for which a perfectly deterministic answer would be unrealistic.

Before Describing What Your Research Has Established, Check:

Frequently Asked Questions About Facts and Uncertainty in Research

Research Can Establish Facts While Also Narrowing What Remains Unknown

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.