What the ITC Requires Before a Cognitive Test Is Translated

What does the ITC require before a cognitive test is translated?

Before adapting a cognitive test for another language and culture, the International Test Commission (ITC) says a team should obtain the rights holder’s permission, assess whether the measured construct and item content fit the target population and intended score use, and identify language or cultural differences that could affect performance for reasons unrelated to that construct. These preconditions do not prove a later version will work equivalently. The ITC uses “adaptation” for the broader work of moving a test between language and cultural settings. Translation carries linguistic meaning across languages; adaptation also considers whether the task, administration, and interpretation remain suitable. The ITC Guidelines for Translating and Adapting Tests organize 18 recommendations into six stages: preconditions, development, empirical confirmation, administration, score interpretation, and documentation. The preconditions come first; later evidence addresses questions that permission or expert judgment alone cannot settle. A team’s first question is not simply who can translate the items. It is whether the work is authorized and meaningful for the people and use in view. If those foundations are missing, translation should wait. A completed translation alone does not establish equivalent measurement or justify comparing scores across language versions.

Sources: The ITC Guidelines for Translating and Adapting Tests (Second edition)

What must be decided before adaptation begins?

The International Test Commission’s three preconditions are permission, construct and content suitability for the target population and intended use, and early identification of cultural or linguistic differences that could affect results for reasons outside the ability assessed. These decisions come before adaptation. They determine whether the project may proceed and what it must examine; they do not certify that a translated form will measure equivalently.

First, the team needs authorization from the test’s intellectual property rights holder. *The ITC Guidelines for Translating and Adapting Tests (Second edition)* recommends a signed agreement before adaptation, clarifying allowed changes and ownership of rights. Access to a test form does not grant permission to reproduce or modify it. If format, materials, or scoring may change, discuss that before work begins. Without authorization, resolve the issue with the rights holder or select an instrument for which permission can be obtained.

Second, ask whether the ability being assessed and its content have enough overlap across source and target populations for the proposed use. A construct is the ability or attribute a test is intended to measure. The ITC recommends involving people familiar with the construct and target groups to judge whether it makes sense in each setting and whether items represent it. This is an early judgment, not proof: the question ultimately needs examination with data. If the ability has a different meaning, or the content does not represent it, translation cannot fix that mismatch. The team may need to reconsider the purpose, choose different content, or stop.

Third, identify language and cultural demands that could affect performance but are irrelevant to the intended ability. The ITC names item format, materials, pictures, computer use, and time limits as areas to consider. Relevance depends on the test and population; a possible issue is not proof of a problem. The review helps decide what deserves attention and whether the task offers a fair opportunity to show the intended ability. For example, a team might check whether an unfamiliar response interface interferes with a reasoning task when interface skill is outside the construct. That is a question to investigate, not evidence that a group will perform differently.

The ITC places these preconditions in a framework of 18 recommendations across six stages, with practice suggestions. It does not prescribe one protocol for every project. The preconditions are a go/no-go point: confirm rights, make a case for construct and content fit, and identify plausible irrelevant demands. If any cannot be addressed, revise the purpose, seek authorized redesign, or choose another instrument before translation. Later empirical work answers different questions about how a developed version functions.

Sources: The ITC Guidelines for Translating and Adapting Tests (Second edition)

Does the ability being measured still make sense in the target setting?

The construct is the ability or attribute a test is intended to measure. Before adapting a cognitive test, reviewers need to ask whether that ability and the test’s content make sense for the people who will take it. The International Test Commission’s *Guidelines for Translating and Adapting Tests (Second edition)* treats this as a precondition: experts who understand both the construct and the target cultural group should judge whether the construct is meaningful in each setting and whether the items adequately represent it. This review helps decide whether adaptation is worth pursuing; it does not establish that two language versions measure the same thing. That distinction matters because matching words is narrower than matching meaning. A sentence may be translated accurately while its premise, context, or expected response draws on knowledge that is unevenly available in the target population. Conversely, an item can be unfamiliar yet still assess the intended ability, depending on the test’s purpose . Familiarity is not automatically a flaw. The question is whether familiarity contributes relevant experience to the construct, or adds an unintended hurdle that changes what a response reflects. Consider an illustrative, invented example: a reasoning task asks someone to infer a rule from a diagram containing a convention used in one school system. If the target group has had little exposure to that convention, difficulty may reflect prior instruction rather than the reasoning process the test is meant to assess. But if the intended construct includes applying learned conventions, removing every unfamiliar reference could also change the content being measured. The example is not a test item or evidence about an assessment. It shows why reviewers must state the intended ability and population before labeling content suitable or unsuitable. The ITC recommends drawing on experts familiar with the construct and target group, and says structured methods such as focus groups, interviews, or surveys can inform judgments about overlap. These are ways to organize evidence and identify concerns, not substitutes for examining how the adapted test performs. The guideline says the question of construct and item-content overlap ultimately needs empirical assessment. In its later confirmation stage, evidence should address whether the test’s structure works equivalently across versions . Analysis depends on the test and intended inference; expert agreement alone cannot settle it. A 2024 neuropsychology-specific application of the ITC framework offers an example. The University of Glasgow record describes recommendations developed by an international working group from literature and specialist expertise to address clinical neuropsychological measures. That application illustrates how a broad framework may need field-specific interpretation; it is not validation of a particular intelligence test, nor proof that its recommendations apply unchanged to every reasoning measure. For a reader evaluating a translated test, look for a documented account of what ability it claims to measure, why the construct and item content fit the target population, and what evidence later supports the intended interpretation. If the documentation offers only a translator’s judgment or a claim that wording is equivalent, the responsible conclusion is limited: the language may have been reviewed, while construct fit and score equivalence remain unestablished.

Sources: The ITC Guidelines for Translating and Adapting Tests (Second edition); Neuropsychological application of the International Test Commission Guidelines for Translation and Adapting of Tests

Which differences can distort a reasoning task without measuring reasoning?

A reasoning task can become harder for a reason the test is not meant to measure. The ITC’s *Guidelines for Translating and Adapting Tests (Second edition)* calls for teams to identify cultural and linguistic differences irrelevant to intended use early in a project. Examples include item format, materials such as computers or pictures, and time limits. The issue is whether it changes what a score represents. A construct is the ability or attribute a test intends to measure. A construct-irrelevant demand is an outside influence on performance. Imagine a person understands a visual relation but loses time navigating an unfamiliar response interface. That demand could affect results when the task measures reasoning. This is an illustration, not evidence of bias in a particular test. Language load is an obvious case to inspect in verbal reasoning. A translated prompt may use uncommon vocabulary, complex syntax, or phrasing that obscures the intended relation. Shortening it may remove necessary information or alter difficulty. Reviewers should distinguish necessary from incidental language demands. Review is relative to the intended ability. Nonverbal tasks can carry demands beyond reasoning. Instructions must be understood; pictures may depict unfamiliar objects; response formats may be unfamiliar; and time limits may interact with test-taking experience. Diagrams may rely on unfamiliar visual conventions. These possibilities do not prove unfairness: timing may be intended if rapid visual search is part of the construct. Connect each suspected demand to a decision. A culturally specific object might be retained if recognition is part of the skill, replaced if recognition is incidental, or examined with target users if its role is uncertain. An interface can be adapted while preserving the reasoning operation, but a changed response method may affect task behavior and should be documented. A time limit should not be removed simply because it is inconvenient if speed is part of the intended interpretation. Before translation, people familiar with the target language, culture, construct, and test use can identify possible method differences. Later, observation, interviews, focus groups, or surveys may reveal whether instructions are understood, materials familiar, or administration as intended. They identify questions for adaptation and empirical study; they do not alone prove equivalent score meaning. A systematic review, *The Effect of Translation and Cultural Adaptations on Diagnostic Accuracy and Test Performance in Dementia Cognitive Screening Tools*, provides a bounded example. Its 17 studies and 4,463 participants showed varied diagnostic sensitivity and specificity across tools; education, language ability, and acculturation also affected performance. The review concerned dementia screening, not general intelligence tests, and included varied tools and studies. It cannot establish causes in IQ tests, but illustrates why “translated” alone is too little information to judge performance in a new setting. Before adaptation proceeds, ask which demands belong to the intended reasoning ability, which may enter through language or context, and what evidence could distinguish them. Identifying plausible sources early lets a team retain, adapt, replace, or study them. A fluent translation cannot show whether a reasoning score reflects the intended ability when access, familiarity, or administration may have changed.

Sources: The ITC Guidelines for Translating and Adapting Tests (Second edition); The Effect of Translation and Cultural Adaptations on Diagnostic Accuracy and Test Performance in Dementia Cognitive Screening Tools

Why does the ITC treat translation as only one stage?

The ITC treats translation as one stage because carrying wording across languages does not, by itself, carry the test’s task demands or justify the same interpretation of its scores. In The ITC Guidelines for Translating and Adapting Tests (Second edition), adaptation is the broader process of deciding whether a test can measure the intended construct in another setting, developing a suitable version, and examining evidence about how that version works. Translation, in the narrower sense used by the guidelines, concerns conveying linguistic meaning. A sentence can be linguistically faithful while the task it presents changes. The distinction matters because the project has several linked decisions. Before development, the team considers whether it has permission, whether the construct and content fit the target population and intended use, and whether irrelevant language or cultural demands may affect responses. Development then creates and reviews a candidate version. This is where decisions about translators, translation design, reconciliation, accommodations, and format changes are made. The ITC describes 18 guidelines organized in six categories: preconditions, test development, confirmation, administration, scoring and interpretation, and documentation. That organization signals a sequence of questions, not six interchangeable checks or a claim that every project follows an identical recipe. Confirmation is a separate step because a well-reviewed draft still needs evidence about its performance. The ITC groups empirical analyses of equivalence, reliability, and validity here. Each kind of evidence answers a different question: reviewers can identify wording or content concerns; empirical analyses can examine whether response patterns and test structure support a proposed interpretation. Neither fluent wording nor a reviewer’s agreement alone establishes that scores have the same meaning across versions. The amount and kind of evidence should follow the use being proposed, including whether scores will be interpreted within one target population or compared across language groups. The remaining categories keep score use connected to the conditions under which the test was adapted. Administration guidance addresses how the test is given; scoring and interpretation guidance concerns how results are derived and used; documentation records the adaptation and evidence so others can judge the basis for those choices. A change in instructions or administration can matter even when item wording looks similar, while documentation makes those decisions visible. The sequence therefore links design choices to the claims made about scores rather than treating translation as a finished product. The Standards for Educational and Psychological Testing, a joint product of AERA, APA, and NCME, offers broader guidance across educational and psychological testing, including fairness and valid score interpretations. It is useful context for responsible testing, while the ITC guidelines address translation and adaptation specifically. Neither a standards title nor a description that a test was “translated” substitutes for evidence tied to the version, population, and intended use. When a provider uses that label, ask which adaptation stages were completed and what each stage supports. If the record describes only linguistic review, the defensible claim is correspondingly narrow: wording was checked. It does not establish suitable score interpretation or cross-language comparability.

Sources: The ITC Guidelines for Translating and Adapting Tests (Second edition); Testing Standards

What can back-translation show—and what can it miss?

Back-translation can reveal a mismatch between a target-language item and its intended meaning, but it answers a narrow question: does a version translated back into the source language appear to preserve the wording? It does not show by itself that target users understand the item as intended, that the item functions similarly, or that scores from two language versions can be compared. The International Test Commission’s *Guidelines for Translating and Adapting Tests (Second edition)* treats translation as one part of adaptation and warns that translation designs alone rarely provide sufficient evidence to validate an adapted test. In back-translation, one translator renders source material into the target language, and another translates that version back into the source language. Comparing this text with the original may expose omissions, additions, or shifts in wording. That can give a team discrepancies to investigate during development. But a close match on the return journey is not the same as a natural, clear, or suitable item in the target language. The process compares texts; it does not test the whole chain from item meaning to response and score interpretation. The ITC identifies a risk when back-translation is used in its narrowest form: the target-language version may receive no direct review. Translators can favor wording that is easy to translate back, even if it sounds awkward to intended readers. Examinees respond to the target-language item they see, not to the source text or its back-translation. If phrasing is unclear, a response may reflect that difficulty alongside the ability the item is intended to assess. This is a reason to inspect the target version, not proof that a particular phrase changes scores. Direct review serves a different purpose. Reviewers familiar with the target language and context can consider whether instructions and items read naturally, retain intended content, and fit the proposed use. They can compare alternative forward translations and reconcile differences in the target language. The ITC describes double translation followed by reconciliation as one approach: independent versions are reviewed by a third translator or expert panel, which resolves discrepancies into a single version. The guideline notes that this allows discrepancies to be considered directly in the target language. It does not require this design in every project or say reconciliation alone establishes measurement equivalence. The useful comparison is between checks with different reach. Back-translation can flag some linguistic discrepancies by returning a target version to the source language. Target-language review can identify problems in the version examinees will actually use. Neither check, alone or together, establishes that the adapted test measures the same construct or supports interchangeable scores. Those are later empirical questions in the ITC framework. A careful report should describe the translation design, whether the target version was reviewed, and how disagreements were handled. Treat back-translation as one possible diagnostic step in documented adaptation, not as a validation result or certificate that scores across languages mean the same thing.

Sources: The ITC Guidelines for Translating and Adapting Tests (Second edition)

What should target-language review and piloting add?

Review and piloting test whether adapted materials work for intended examinees. Review can identify problems before administration; a pilot shows how instructions and procedures function in use; empirical confirmation examines properties observation cannot establish. None alone certifies interchangeable scores across language versions. The ITC Guidelines for Translating and Adapting Tests place development work after the project’s preconditions and before empirical confirmation. Reviewers need relevant knowledge of language, target culture, test content, and assessment principles. The ITC allows expertise to be distributed across a team; the record should show what informed decisions. Review should examine meaning and function, not just whether each sentence has a counterpart. Reviewers can check whether instructions are clear, vocabulary comparably familiar, wording changes difficulty, and response choices still work. They can inspect timing and administration materials. The aim is not identical wording across languages. Reviewers should notice when a wording or format change alters the task and decide whether it preserves the intended demand or needs study. The ITC document describes reviewer judgment as useful for checking these features and discusses reconciliation that compares alternative target-language versions directly. A pilot adds evidence because members of the intended population encounter the candidate materials. Their questions or misunderstandings may show that wording reviewers considered clear is interpreted differently in practice. Feedback can reveal problems with instructions, timing, or procedure. These observations are diagnostic: they help locate a possible source of difficulty and guide revision. They do not establish how common the problem is, whether it changes scores systematically, or whether the test measures the intended construct adequately. The 2024 article “Neuropsychological application of the International Test Commission guidelines for translating and adapting of tests” shows the role of field context. Its working group developed recommendations specifically for neuropsychological tests, drawing on literature and international expertise. The University of Glasgow record describes a specialized application of general ITC guidance, not validation of any particular translated instrument. A general framework may need domain-specific interpretation; evidence for a version must match its intended use. Teams should document changes and reasons, including disagreements affecting wording or procedure. A revision record identifies the version piloted and helps evaluators follow decisions. The ITC treats documentation as a separate stage. A useful account names the target population, reviewers’ roles, pilot conditions, observed issues, and resulting changes. This record makes reasoning inspectable; it does not substitute for evidence. The next question is whether the revised form supports the proposed interpretation. That requires empirical analyses matched to the claim, such as item behavior, test structure, reliability, or validity evidence. The ITC separates confirmation from development for this reason. Review and pilot work can improve a version and identify questions to test, but acceptance or smooth administration cannot establish score precision, construct equivalence, or fairness. A credible account says who reviewed and tried the materials, what changed, and which analyses support the proposed use.

Sources: The ITC Guidelines for Translating and Adapting Tests (Second edition); Neuropsychological application of the International Test Commission Guidelines for Translation and Adapting of Tests

What do item analyses establish?

Item analysis can flag a question whose response pattern differs between language groups after accounting for the ability the test is intended to measure. That signal can guide review, but it does not explain the difference or establish that total scores from the two versions are comparable. The ITC Guidelines for Translating and Adapting Tests (Second edition) places item equivalence within a wider empirical examination that also considers construct and method equivalence. Item statistics contribute evidence; they do not settle the whole question of score meaning. Differential item functioning (DIF) means a difference in the probability of a particular response between groups whose members are matched on the measured trait. For example, among examinees at a similar level of the ability under study, one language group might be more likely to answer a particular item correctly. This conditional item-level comparison differs from finding that one group has a higher average score on the full test. A group average alone does not show that an individual item behaves differently for otherwise similar examinees. The ITC guideline distinguishes item-level DIF from differences in overall group scores. A flagged item is a prompt for investigation, not a verdict. Reviewers need to examine what the item asks people to do and how they may understand or answer it. Wording might carry an unintended meaning; a diagram, response format, or instruction might introduce a demand outside the intended cognitive ability. These are possibilities to examine, not causes proven by the statistical signal. The ITC Guidelines for Translating and Adapting Tests (Second edition) treats linguistic and cultural influences, test method, item equivalence, and construct equivalence as related but distinct concerns. The item’s content and adaptation record can help frame interpretation; evidence about response processes or administration may also be relevant to the proposed use. The analysis must fit the question. A method appropriate for one response format, sample, and comparison may not answer another. The ITC does not set one DIF statistic as a universal requirement for every cognitive test adaptation. Detection alone also does not show whether a difference materially affects score interpretation. Investigators should consider the method’s assumptions, the relevance of the data, and whether the analysis matches the versions, populations, and intended inference. Item-level evidence cannot replace evidence about broader test structure, reliability, validity, or norms. When reviewing an adapted test, ask which groups and versions were compared, how examinees were matched on the measured trait, what analysis was used and why, which items were flagged, and how the signals were investigated. Then ask whether the results support the proposed interpretation, especially if scores are to be compared across languages. A flagged item does not prove that a group or individual is disadvantaged; no reported flag does not by itself certify equivalence. Item analysis can identify where closer review is needed, while the wider evidence must support the conclusion.

Sources: The ITC Guidelines for Translating and Adapting Tests (Second edition); PubMed record on differential item functioning

When can the original norms or score scale carry over?

Original norms can carry over only when evidence supports using them fairly for the adapted test, target population, and intended interpretation. A norm is a reference distribution for performance. It differs from reliability, about score consistency; validity evidence supports a particular use. Sharing a score scale does not establish equivalence between language versions. The ITC Guidelines for Translating and Adapting Tests (Second edition) state that a project proposing the source version’s norms should show their use is statistically appropriate and fair. Without such evidence, the guideline says to develop specific norms for the adapted version. This is conditional: every translation does not automatically require new norms, but the group and interpretation must be justified.

The evidence needed depends on the claim. A team may ask whether the target-language form supports an intended use there. A different claim is that scores between language versions can be compared directly or treated as interchangeable. The ITC notes that equivalence between versions is not always the goal. For use within a different language group, validity evidence for that group may be essential even when direct equivalence research is less critical. Comparison across groups requires evidence for that inference; linking scales also needs an appropriate design and analysis. Local usefulness alone does not justify cross-language rankings or score conversion.

Ask whether the adapted version supports local interpretation, its norms suit that population, and direct comparison is supported. Reliability evidence concerns consistency, not norm-group fit. One supported interpretation does not establish another, and identical labels do not create a sound score link. Documentation should state population, use, and evidence.

The systematic review Reporting of the translation and cultural adaptation procedures of the Addenbrooke’s Cognitive Examination version III (ACE-III) and its predecessors shows why process records alone cannot establish measurement quality. It reviewed 32 ACE-family publications and assessed reporting of adaptation procedures; its authors caution that reporting measures could not determine whether an assessment was appropriately adapted. These are cognitive screening instruments, not general intelligence tests, so this is an illustration, not evidence about every test. Accept original norms or a shared scale only for the population and use supported by evidence. Translation alone does not establish comparability.

Sources: The ITC Guidelines for Translating and Adapting Tests (Second edition); Reporting of the translation and cultural adaptation procedures of the Addenbrooke’s Cognitive Examination version III (ACE-III) and its predecessors; The Effect of Translation and Cultural Adaptations on Diagnostic Accuracy and Test Performance in Dementia Cognitive Screening Tools

What should a user ask for before trusting the result?

Before trusting an adapted cognitive test, ask what decision the score is meant to support. Local use makes a different claim from comparison across language versions. The International Test Commission Guidelines for Translating and Adapting Tests make that distinction consequential: evidence should match the proposed interpretation, and equivalence between versions is not always the project’s goal. If direct comparison is intended, ask for evidence that addresses it rather than assuming careful translation supplies it. Then ask whether the work was authorized and whether the target population and construct were examined before translation began. Request an account of the adaptation process: who reviewed content for the target setting, whether people from the intended population tried the materials, what problems appeared, and how revisions were handled. Ask what empirical confirmation was conducted and what it can support. The ITC provides guidance, not one protocol or sample size for every test. Finally, separate local interpretation, norm-referenced interpretation, and cross-language comparison. Norms are a reference group used to interpret performance; they do not by themselves show that two versions measure equivalently. Ask whether the source norms were shown to suit this version and population, or whether other norms were developed, and what evidence supports the intended use. The AERA, APA, and NCME Standards for Educational and Psychological Testing offer broader guidance on fair testing and valid score interpretation; the ITC addresses adaptation specifically. If the record documents only linguistic checks, keep the conclusion narrow: wording was examined, but suitable norms or interchangeable scores have not been established. Proceed only with the supported use; when a consequential decision exceeds that support, seek a qualified assessment suited to it.

Sources: The ITC Guidelines for Translating and Adapting Tests (Second edition); Testing Standards