Should You Restrict a Literature Search by Language?

Should You Search Only for Studies Published in a Language You Can Read?

You search a database and retrieve papers in English, Chinese, Spanish, German, Japanese, Portuguese, and several other languages. You can comfortably evaluate only the English-language records.

The obvious solution seems to be an English-language filter.

For some projects, especially rapid or resource-constrained reviews, a language restriction may ultimately be necessary. But the decision is not methodologically neutral. If relevant studies are published in languages other than the one you select, restricting by language can change which evidence has an opportunity to enter the review.

The issue is therefore not simply whether you can personally read every language represented in the literature. You need to consider whether excluding studies by language could systematically alter the evidence you find, whether the topic has important research outside English-language sources, and whether translation or other practical alternatives are available.

Avoid Language Restrictions When Comprehensive Evidence Identification Matters

Language Is More Than a Practical Reading Problem

A language restriction determines which evidence can be seen

A language filter may look like a simple database convenience. Methodologically, however, it is an exclusion mechanism.

Suppose a database contains 800 records potentially relevant to your question. Seventy are published in languages other than English. If you apply an English-only filter, those records never reach screening.

You therefore cannot determine whether they were eligible based on their substantive content because the search removed them first.

This distinction matters because language metadata can be missing, inconsistent, or incorrect. Cochrane's current guidance consequently recommends that systematic review searches not be restricted by language and notes that, when a language restriction is justified, applying it during study selection rather than as a database search limit can avoid some problems associated with metadata-based filtering.

Language restriction can introduce language bias

The concern is not merely that you will retrieve fewer studies. The excluded studies may differ systematically from those that remain.

Cochrane identifies language bias as a potential consequence of restricting searches according to publication language. Research findings, publication decisions, journal selection, and patterns of international dissemination can interact with language in ways that make the retained evidence an imperfect representation of all available evidence.

This is one reason comprehensive systematic reviews seek to identify eligible studies irrespective of publication language.

The effect of an English-only restriction is not identical for every topic

There is an important nuance here.

Empirical studies examining English-language restrictions have often found that excluding non-English reports does not substantially change the conclusions of many systematic reviews. Cochrane cites evidence supporting this observation, including research on rapid reviews and meta-analyses.

But that is not equivalent to saying language restrictions are harmless.

Cochrane also identifies areas in which non-English studies have influenced review findings, including complementary and alternative medicine and some work in psychiatry, rheumatology, and orthopaedics. It further warns that English-only searching can be particularly problematic when an intervention is common in countries whose research is frequently published in other languages.

The methodological question is therefore not “Does language restriction always cause bias?” It is “What evidence might become systematically unavailable if I impose this restriction in this review?”

Searching an English-language database without a language filter is not the same as searching globally

This distinction is easy to miss.

You can remove the English-language filter from a database and still fail to identify substantial non-English literature.

Cochrane explicitly notes that removing language restrictions in English-language databases is not a substitute for searching relevant non-English-language journals, databases, and trials registers. A database may disproportionately index English-language or internationally visible journals even when it contains some records in other languages.

A systematic review can therefore truthfully report “no language restrictions” while still having limited access to research published outside the databases it searched.

This is one reason the decision about how many databases you need to search cannot be separated entirely from questions of language, geography, and disciplinary coverage.

Your search terms can create a language restriction even when your filters do not

Suppose you do not activate an English-language filter, but every free-text search term in your strategy is English.

Have you conducted a language-unrestricted search?

Technically, the database may be allowed to return records in any language. In practice, retrieval depends on how those records are indexed and what searchable metadata they contain.

Some non-English articles have English-language titles, abstracts, keywords, or controlled vocabulary and may therefore be retrievable using English terms. Others may not.

Cochrane consequently recommends considering databases and trials registers published in languages other than English when warranted and involving information specialists capable of designing and executing searches in those sources.

Language coverage is therefore partly a database-selection problem and partly a search-strategy problem.

Controlled vocabulary can help, but it does not eliminate the issue

Databases that apply standardized subject indexing can sometimes help bridge linguistic variation. A non-English report may be assigned controlled vocabulary that can be retrieved using the database's standardized subject headings.

That is useful, but it does not guarantee complete multilingual retrieval.

Not every database uses the same controlled vocabulary. Not every record is indexed to the same depth. Newer records may have incomplete indexing. Some databases rely heavily on author-supplied metadata.

A comprehensive search should therefore consider both free-text terminology and appropriate controlled vocabulary while recognizing that database coverage still constrains what can be found.

Language can be part of eligibility without being imposed during searching

Suppose your project genuinely cannot analyze publications in languages other than English. One approach is to apply an English-only database filter immediately.

Another is to retrieve records without that filter, identify potentially relevant non-English reports, and then apply the pre-specified language eligibility rule during study selection.

The latter preserves information about what evidence was encountered and avoids relying entirely on database language metadata. Cochrane cites methodological arguments recommending this approach when language restrictions are genuinely necessary.

This follows the broader distinction between eligibility criteria and search restrictions. Not every characteristic that determines final inclusion needs to be used to suppress records during retrieval.

A language restriction should have a rationale more specific than convenience

Research projects have finite time, money, personnel, and translation capacity. Those constraints are real.

But “we only speak English” describes a resource limitation, not evidence that non-English studies are irrelevant.

If a restriction is unavoidable, say what it is and why it was imposed. JBI guidance similarly requires language restrictions to be stated and expects search timeframes and restrictions to be justified within the relevant review methodology.

A transparent statement allows readers to interpret the evidence base accordingly.

The wording should, of course, reflect what you actually did. Transparency is more useful than pretending that a practical constraint is a substantive characteristic of the research question.

Rapid reviews may legitimately make different trade-offs

Comprehensive systematic reviews and rapid reviews do not always use identical search methods.

Cochrane's current Handbook notes that guidance for rapid reviews may permit restricting publication language to English, with additional languages included when justified. Evidence suggests that English-only restrictions do not alter the conclusions of many reviews, although exceptions exist.

This is a methodological trade-off. A rapid review intentionally accepts some loss of comprehensiveness in exchange for faster production.

The key is to describe the review accurately. A search that intentionally uses streamlined methods should not be presented as though it had the same evidence-identification scope as an unrestricted comprehensive systematic review.

Do not assume that translation requires professional translation of every paper

Encountering a non-English record does not necessarily mean commissioning a complete certified translation.

Depending on the purpose and methodological requirements, researchers may be able to use:

• a multilingual member of the research team;
• a colleague or collaborator who reads the language;
• institutional translation support;
• research networks or volunteer assistance;
• machine translation as an aid to initial understanding, with appropriate human verification where decisions or extracted data require greater confidence.

Cochrane specifically recommends seeking assistance with translation or data extraction when required for reports published in languages other than English.

The appropriate method depends on what must be extracted and how consequential an error would be. Understanding whether an abstract is potentially eligible is a different task from extracting numerical outcome data for a meta-analysis.

Machine translation can reduce practical barriers, but it does not erase methodological responsibility

Modern translation systems make multilingual literature more accessible than it once was. They can be useful for preliminary screening, understanding abstracts, identifying terminology, and supporting communication with multilingual collaborators.

However, automated translation can misinterpret technical terminology, negation, statistical language, culturally specific expressions, or complex methodological descriptions. Its reliability also varies by language pair and source text.

For consequential decisions, especially eligibility judgments or quantitative data extraction, researchers should verify translated information to a level appropriate to the risk of error.

A machine-translated article does not become methodologically self-validating merely because the English output reads fluently.

Do not automatically exclude a study because you cannot immediately obtain a translation

For a systematic review, there is an important difference between determining that a study is ineligible and being temporarily unable to determine its eligibility.

Cochrane advises that where relevant information cannot be extracted from reports in languages other than English, potentially relevant reports should be identified transparently rather than simply classified as excluded. In Cochrane reviews, such reports may be placed among studies awaiting classification when eligibility cannot yet be determined.

This distinction prevents lack of translation capacity from being disguised as substantive ineligibility.

Language restriction can matter especially for geographically specific questions

Suppose your review concerns educational technology implementation in Southeast Asia. Searching only English-language international journals could omit studies published in national journals, university repositories, government reports, or regional databases.

The same issue can arise in health, agriculture, environmental research, social policy, and other fields where local evidence is particularly important.

A geographically focused question should therefore prompt you to ask which languages researchers, governments, institutions, and journals in that setting actually use.

This may influence database selection as much as the language filter itself.

Language and publication status can interact

Relevant non-English evidence may be disproportionately represented outside the major commercial journal databases, including theses, reports, conference proceedings, local journals, institutional repositories, and other grey literature.

If you restrict simultaneously by language and publication type, the combined effect can be more substantial than either restriction considered alone.

This is why the question of whether to search only peer-reviewed literature deserves separate consideration. A review can become narrower through several individually convenient decisions that collectively remove entire segments of the evidence base.

Search restrictions should match the claims you intend to make

If you search only English-language publications, you can still conduct useful research. The problem arises when the conclusions are framed as though language could not have affected the evidence identified.

Consider the difference between these claims:

The available evidence shows...

and:

Among the English-language literature identified by this review...

The second formulation acknowledges the boundary when that boundary matters.

The broader principle is simple: the scope of your conclusion should not silently exceed the scope of your evidence.

Document what you actually did

For reproducible searching, record whether language restrictions were used, where they were applied, and what languages were eligible.

If no language restriction was applied, record that too when relevant to the review methodology. If potentially eligible non-English reports required translation or remained unresolved, preserve that information in the study-selection record.

Do not rely on memory. Months later, “I think we searched everything” is not quite the methods section reviewers dream of.

The broader documentation needed to make the search reproducible is addressed in what you should record during a literature search.

What Should You Do When Relevant Studies Appear in Languages You Cannot Read?

The practical workload is greater than simply clicking “English.” But the team can now distinguish evidence that was genuinely ineligible from evidence that would otherwise have disappeared because nobody on the original team happened to speak its publication language.

Common Mistakes About Language Restrictions

Separate What Is Methodologically Desirable From What Is Practically Possible

If your goal is comprehensive evidence identification, begin from the position that potentially eligible evidence should not be discarded merely because of publication language.

Then consider the actual linguistic demands of your topic and the resources available to address them.

For less formal literature searching, the same principle can be scaled to the purpose of the project. You may reasonably decide to work only with literature you can evaluate reliably. Just avoid turning that practical boundary into an unsupported claim that relevant research in other languages does not exist.

Before Restricting Your Literature Search by Language

Questions About Language Restrictions in Literature Searches

Do Not Let Language Quietly Decide Which Evidence Counts

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.