Replication vs. Extension Study: What’s the Difference?

Are You Testing the Previous Claim Again or Moving Beyond It?

You find a published study and decide to build on it. You retain much of the original design, but perhaps you add another variable, include a different outcome, collect data from another group, examine a longer period, or test an additional hypothesis.

Is that still a replication, or have you created an extension study?

The distinction matters because the two make different kinds of contributions. Replication primarily generates new evidence relevant to something already claimed or observed. Extension moves beyond the original investigation by asking something additional. Yet the categories are not mutually exclusive, and terminology varies across disciplines. A well-designed project can contain both replication and extension components.

The Difference Depends on What the New Study Is Trying to Establish

Replication and Extension Differ Most Clearly in Their Inferential Purpose

A Replication Looks Back at an Existing Claim

Replication begins with prior evidence. The researcher identifies a finding or claim from an earlier study and collects new evidence that can change how confidently that claim should be interpreted.

Nosek and Errington frame replication in terms of whether a new study provides evidence that is diagnostic of a claim from prior research. This is more useful than defining replication simply as copying procedures. A replication may resemble the original study very closely, or it may differ in meaningful ways, depending on what claim is being tested.

The central logic is nevertheless retrospective:

Previous research made or supported this claim. What happens when we test that claim again?

This can involve a close methodological repetition or a deliberately varied test. The latter may be described as conceptual replication in some disciplines, although direct and conceptual replication address somewhat different evidential questions.

An Extension Moves Beyond What the Original Study Established

An extension begins from previous research but deliberately expands the inquiry. Rather than focusing only on whether an earlier result recurs, it asks something that the original study did not answer.

For example, an extension might investigate:

• a theoretically meaningful moderator that was not examined previously;
• a possible mechanism that could explain the original relationship;
• an additional outcome or consequence;
• a longer time horizon;
• a new condition under which the relationship might strengthen, weaken, or reverse;
• an additional comparison not included in the original design; or
• a new hypothesis derived from the original findings.

The defining feature is not merely that something has changed. Something substantive is being learned beyond the original inferential target.

A Methodological Change Does Not Automatically Create an Extension

This is where the distinction becomes less tidy.

Suppose the original study used one measure of academic engagement and you use another defensible measure of the same construct. You have changed the method, but you may still be testing essentially the same underlying claim. Depending on the disciplinary taxonomy, that could be considered a conceptual replication rather than an extension.

Likewise, conducting the study in another setting does not necessarily extend the theoretical claim. If your purpose is to determine whether the same finding occurs under another condition, the project may still primarily function as a replication.

This is why counting differences between the two studies is a poor classification strategy. One changed feature can create a substantial extension if it introduces a genuinely new inferential question, while several methodological changes can still be directed toward testing essentially the same underlying claim.

Changing the Population Can Be Replication, Extension, or Both

Population changes are especially difficult to classify without knowing why they were made.

Imagine that an original study reports a relationship between academic self-efficacy and persistence among undergraduate students. You repeat the study among another sample of undergraduate students from a comparable population. The new sample primarily provides another test of the existing relationship.

Now suppose you deliberately study working adult learners because theory suggests that employment demands may alter the relationship. If you test only whether the original relationship appears in this population, you are still generating replication evidence about its generalizability. If you additionally test whether employment demands explain differences in that relationship, you have introduced an extension question.

The same population change can therefore serve different purposes depending on the claims built into the design. Determining what changing the population actually changes about the research question is more informative than treating any new sample as proof of novelty.

Adding a Variable Does Not Automatically Make the Study an Extension Either

A similar problem occurs when researchers add variables.

Suppose the original study examines whether perceived instructor support predicts student engagement. You repeat that analysis but also measure students' year level because you need it to describe the sample or use it as a justified adjustment variable. The mere presence of an additional measured variable does not necessarily create an extension.

Now suppose you add academic self-efficacy and hypothesize that it mediates the relationship between instructor support and engagement. That addition introduces a new explanatory claim. Your study is no longer only asking whether the original relationship recurs. It is also asking how that relationship might operate.

The important issue is therefore not the number of variables but their inferential role. This distinction becomes particularly useful when determining when adding a new variable actually turns replication into extension.

Replication and Extension Can Occur in the Same Study

You do not always have to choose one label.

Bonett describes replication-extension studies as designs that combine results from prior studies with a new study specifically designed to replicate and extend the earlier findings. Such designs can assess replication while also addressing questions about effect-size precision, generalizability, misleading prior findings, or moderator effects.

Consider a study that first reproduces the primary analysis from an earlier investigation and then tests an additional theoretically motivated moderator. The first analysis provides replication evidence. The second extends the original inquiry.

This combined design can be useful because it preserves a clear test of prior evidence while using the same project to investigate a justified next question. Empirical examples across fields have explicitly combined replication with extended data, analyses, or time frames.

Do Not Let the Extension Contaminate the Replication Test

Combining replication and extension has an important design implication: you should be able to distinguish the replication component from the extension component.

Suppose the original study tested whether X predicts Y. You collect X and Y but also add Z as a proposed moderator. If your replication test now consists only of a new model containing X, Y, Z, and an interaction term, you may no longer have a clean comparison with the original finding.

A clearer approach may be to distinguish the analyses:

The exact analytical strategy depends on the research design and statistical model, but the underlying principle is straightforward. Readers should be able to tell which evidence speaks to the previous claim and which evidence speaks to the new one.

Extension Is Not a Device for Making Replication Look More Novel

Researchers sometimes feel pressure to add something to a replication because retesting the original claim appears insufficiently novel. This can produce what is essentially methodological decoration: another variable, another subgroup, another outcome, perhaps another analysis, all added primarily so the study can claim originality.

That is a weak reason to extend a study.

An extension should follow from a substantive uncertainty. If you add a moderator, there should be a reason to expect the original relationship to differ across levels of that moderator. If you add an outcome, explain why that outcome advances understanding. If you extend the observation period, explain what the longer period could reveal that the original time frame could not.

Terminology Is Not Fully Standardized Across Research Fields

You may encounter papers using replication and extension differently from the definitions above. That is not necessarily an error.

A broad review of replication research found substantial conceptual variation across scientific fields, including definitions that treat replication as repetition, extension, or a form of theory testing. Earlier methodological discussions likewise show that researchers have drawn the replication-extension boundary in different places.

For example, what one tradition calls a conceptual replication may resemble what another calls a replication with extension. Some researchers use "extension" when the population, variables, or design change while the original conceptual relationship remains intact. Others reserve extension for studies that introduce a genuinely additional hypothesis.

Consequently, classification should not become a semantic contest. Define what you mean and describe the relationship between your study and the original research precisely.

The Best Test Is to Separate the Claims

If you are unsure whether your project is a replication, extension, or both, write down every major claim the study is designed to evaluate.

Then ask whether each claim was already examined in the original study.

This claim-based approach is often more defensible than deciding based on superficial similarity between methods. It also helps answer the broader question of whether you should replicate an existing study or pursue a genuinely different research question.

One Project Can Replicate a Finding and Extend the Research

If the researcher measured feedback literacy only as a descriptive characteristic, however, its presence would not automatically make the project an extension. The extension arises because the variable is used to test an additional substantive claim.

Likewise, if the researcher simply replaced the original writing assessment with another defensible operationalization of writing performance while still testing the same underlying claim, that methodological change might be better understood as part of a conceptual replication rather than an extension.

The classification therefore depends on what the new design allows the researcher to infer, not on how many visible differences can be found between the two methods sections.

Common Mistakes When Distinguishing Replication From Extension

Describe What Your Study Retests and What It Adds

When planning a study based closely on previous research, begin by writing the original study's central claim in one sentence. Then write your proposed claims underneath it.

This simple exercise usually reveals the structure of the project.

Your title, abstract, introduction, hypotheses, methods, and discussion should preserve this distinction. If the study has two purposes, say so. Do not present extension findings as though they were evidence that the original result replicated, and do not bury an unsuccessful replication beneath a more interesting extension result.

This is especially important when the replication and extension components produce different patterns. You might obtain evidence consistent with the original relationship but no evidence for your proposed moderator. Conversely, the original relationship may not recur while an exploratory extension produces an interesting association. Those outcomes answer different questions and should be reported as such.

A transparent replication does not need to disguise itself as unprecedented research. The stronger approach is to justify the replication for the uncertainty it addresses and justify the extension for the additional knowledge it is designed to produce.

Before Calling Your Study a Replication, Extension, or Both

Questions About Replication and Extension Studies

Separate What You Are Retesting From What You Are Adding

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.