Translating results without translating away uncertainty
SN-2026-092 · Practice essay · Communities and governance
By Kaisa Lehtinen · · 3 pages · 1,539 words
Fictional demonstration publication. Not peer reviewed. Still North Arctic Institute is a fictional organisation. This paper is a teaching synthesis or illustrative practice document, not a report of original institute fieldwork. Published sources are identified below; illustrative examples are not measured results.
Abstract
Multilingual return of results must preserve the relationship between evidence and claim, not merely reproduce sentence structure. This practice essay compares research-governance sources with W3C accessibility guidance through a fictional seasonal-observation case in which an ambiguous denominator changes the apparent conclusion. It examines qualified translation, technical review, terminology, counts, missing records, chart explanation, confidentiality, and synchronised corrections across written and spoken formats. The discussion distinguishes accessibility from translation quality and separates review roles from endorsement. No actual translations, community consultations, or reader evaluations are reported. The essay rejects treating English as automatically authoritative and offers responsibilities for outside institutions without prescribing a universal Indigenous format. Its central implication is to resource multilingual return as part of evidential integrity throughout a project's lifetime.
Download the complete PDF (3 pages)
Full paper
Return is a design obligation, not a translation receipt
A research summary can be translated accurately word by word while changing the conclusion a reader takes away. A term equivalent to “not detected” may be understood as “not present”; a probability may become a declaration; a regional average may sound like a statement about a particular place. Multilingual return of results therefore requires attention to the scientific claim, the communicative setting, and who has authority over representation.
This practice essay compares research-governance sources with W3C guidance on accessible explanation. It develops a fictional design case, not an evaluation of an actual return programme. It does not provide translations into an Indigenous language, claim community approval, or prescribe a universal Indigenous communication format. The question is what an outside research institution should preserve and resource when agreed results must remain meaningful across languages and media.
Agree the relationship before selecting the format
The National Inuit Strategy on Research connects dissemination with Inuit governance, ownership, control, access, capacity, and language. Its priorities concern Inuit Nunangat research, not a generic audience segment. An outside institution cannot infer a community's preferred language, dialect, spokesperson, or review process from that national strategy. Those decisions require the relevant authorities, knowledge holders, and intended readers.
TCPS 2 Chapter 9 addresses opportunities for community representatives in collaborative research to participate in interpretation and review before final reporting. It also recognises diversity within communities. One meeting, one translator, or one representative's comments cannot honestly be reported as unanimous endorsement. Scientific review, cultural review, permission to attribute, and approval to distribute may be different decisions held by different people.
A return plan should make those responsibilities explicit without forcing them into an imported template. Relevant questions include who will use the results, who may receive particular details, whether oral and written accounts serve different purposes, and what routes exist for correction. The institution proposing the work should resource the agreed processes rather than expect participants to perform translation and review as unpaid proof of engagement.
Fictional case: the denominator changes in translation
Imagine an invented seasonal observation project that visited 12 sites in an agreed study area. Eight sites supplied usable records for a specified comparison; four did not. The analysis finds a particular pattern in five of the eight usable records. These counts are teaching examples, not findings about any real place. The proposed return package includes a concise written account, a chart with a text equivalent, and an optional spoken explanation if that format is wanted.
The English draft says, “The pattern occurred at most sites.” This is already ambiguous before translation. Most of the 12 visited sites, most of the eight analysable sites, or most sites across the wider area? A translation team cannot repair that scientific ambiguity by choosing a more fluent phrase. The analytical team must first specify the denominator and the limits of inference.
A more exact source proposition is: five of eight sites with usable records showed the defined pattern; four visited sites lacked usable records; the design does not estimate the fraction of all sites in the wider area. That proposition may need several sentences, a spoken example, or a different explanatory sequence in another language. Fidelity concerns the relationship between evidence and claim, not identical sentence structure.
| Meaning to preserve | Potential distortion | Design response in the fictional case |
|---|---|---|
| Five of eight usable records | Five of all visited sites, or most of the region | State both the analysed denominator and the missing records |
| No usable measurement at four sites | No change occurred at those sites | Keep missingness separate from an observed negative result |
| A comparison within a defined season | A permanent condition | Keep the period beside every summary form |
| An interpretation with limits | An endorsed community position | Separate analytical attribution from participation and approval |
Preserve a set of claims, not an untouchable English draft
The team can maintain a concise account of the essential propositions: question, population, period, measured quantity, comparison, uncertainty, and limit. Each language version should express these coherently in its own form. English should not automatically be treated as the scientifically authoritative version while all others are derivative publicity. Scientific ambiguity discovered through translation is feedback on the analysis and writing, not a failure of the receiving language.
Technical terms may lack a simple equivalent or may carry different ordinary meanings. Translators and relevant language experts need access to the underlying explanation, not only a glossary of isolated nouns. A definition or example may be preferable to an unfamiliar borrowed word. Where an equivalent remains unsettled, state that difficulty and agree how it will be represented rather than silently substituting a more certain or more alarming term.
Back-translation can reveal discrepancies, but agreement between two English versions is not sufficient evidence that intended readers understand the result. It may reward literal phrasing that is awkward or misleading in use. A stronger process combines qualified translation with review of technical meaning and, where agreed, a reader task that asks someone to explain the finding and its main limit. This is a proposed design approach, not a claim that such testing occurred.
Numbers need context that travels
Fractions, percentages, ranges, and probability words need explanations that remain consistent across formats. In the fictional case, five of eight is 62.5%, but that precise percentage can draw attention away from the small, selected denominator. A short account might lead with the count, while a technical table preserves the numerical detail. Neither form should erase the four unusable records.
Uncertainty is not interchangeable with indecision. A sentence can state a well-supported measured pattern clearly while explaining that spatial generalisation is unsupported. Repeatedly inserting a vague equivalent of “maybe” can make every part of the result sound equally doubtful. Put the limit on the specific inference it constrains: the study area was not sampled in a way that establishes a regional proportion.
Charts create additional translation work. Units, legend terms, axis notes, missing-value categories, and source lines all carry meaning. W3C's complex-image guidance recommends text that conveys scales, values, relationships, and trends, not merely that a graph exists. A translated summary above an untranslated chart is an incomplete return product, and a table alone may not explain why the comparison matters.
Accessibility and language are related, not equivalent
A language version can remain inaccessible because of reading load, visual design, audio quality, disability, connectivity, or file size. Spoken delivery may help one audience and exclude another. Print may be valuable where connectivity is unreliable. A compact web page may be easier to navigate than a large PDF. These are design choices to establish with intended users rather than assumptions attached to a language or community.
WCAG 2.2 provides a web-accessibility standard, but it is not a translation-quality certification or a guarantee of cultural relevance. Technical accessibility and communicative suitability need separate attention. A technically conformant page can still misstate the study denominator, and a fluent spoken account can still omit the route to the evidence.
Interpreters also encounter confidential information. TCPS 2 Chapter 3 notes both the need for appropriate language skills in consent communication and possible confidentiality issues when intermediaries are involved. Returning results is not identical to obtaining consent, but the practical distinction remains relevant: someone should not have to disclose a private question in a public meeting or through an intermediary they have not agreed to use.
Corrections must reach every version
A multilingual package needs a shared release identity and a record of which versions express the same analysis. If the denominator is corrected from nine usable sites to eight, updating only the English web page leaves an older conclusion circulating elsewhere. The correction plan should cover print copies where feasible, downloadable files, audio, captions, and notices through the agreed distribution routes.
Record review roles accurately. A translator's work is not evidence of scientific endorsement; a participant's correction of a place name is not approval of every interpretation. Permission to name a reviewer or quote feedback should be obtained separately. Reports can describe the review process without inventing satisfaction measures or claiming that silence meant agreement.
Limits and implication
The fictional case tests a narrow denominator problem. It does not represent the complexity of particular languages, knowledge systems, or governance arrangements, and no actual readers were consulted. The sources do not specify one appropriate medium, review sequence, or terminology for every project. Material findings that require qualified assessment also need an agreed communication route; translation work should not be used to improvise technical advice.
The implication is to make multilingual return part of the research's evidential integrity. Resource qualified work, let linguistic review improve the source explanation, preserve the claim and its limits across formats, and maintain corrections together. A result has been returned usefully when the intended reader can understand what it answers and what it cannot answer, without being made to endorse the institution that delivered it.
References
- Inuit Tapiriit Kanatami (2018). National Inuit Strategy on Research.
- Canadian research agencies (2022). TCPS 2, Chapter 9: Research Involving the First Nations, Inuit, and Métis Peoples of Canada.
- Canadian research agencies (2022). TCPS 2, Chapter 3: The Consent Process.
- World Wide Web Consortium (2024). Web Content Accessibility Guidelines 2.2.
- W3C Web Accessibility Initiative. Complex Images: Images Tutorial.
Keywords: multilingual communication; return of results; accessible evidence; translation review.
Browse and filter the research library · Download learning guides