Found in Translation roundup – April 2026
Welcome to this month’s Found in Translation roundup!
Found in Translation is a fun little thing I do over on Instagram where once a week I take some of the most commonly asked questions about translation and try to answer them. It can be anything really, from abstract arguments about different understandings of translation quality to practical tips about how to approach your first translation project with a translator.
But not everyone is on Instagram. And even if they are, it’s easy to miss something, so here for convenience is a monthly roundup of all the post from this month, along with a brief summary. Enjoy!
What is the difference between translation and interpreting?
Translation and interpreting are often mixed up, but the traditional distinction is straightforward: translation is written-to-written work, while interpreting is spoken-to-spoken (for example, translating a book versus interpreting a speech). In real professional practice, however, language work frequently crosses these boundaries. Subtitling, usually grouped under “audio-visual translation,” illustrates this complexity because it typically converts spoken dialogue into written text in another language, meaning the practitioner shifts not only between languages but also between media. This creates practical constraints: because most people read more slowly than they can understand speech, subtitlers cannot reproduce every word. Instead, they use “condensing,” making informed choices about what information is essential and how to present it so subtitles remain readable and comfortable to follow. The movement between written and spoken modes also happens in interpreting through “sight translation,” where an interpreter reads a written text and delivers a spontaneous spoken version without drafting it first. Audio-visual work can also involve questions around accessibility, such as subtitling for the deaf and hard-of-hearing (SDH), which may need to indicate who is speaking, tone of voice, and relevant non-speech sounds. Overall, many linguistic roles require additional skills like contextual judgement, prioritising, distilling, and adapting content for the intended user.
How many languages do translators speak?
Translators are often assumed to speak many languages, but the basic requirement is simple: at least two – one source language (the language you translate from) and one target language (the language you translate into). Most translators work mainly into a single target language because best practice is to translate into your native language (or “language of habitual use”), where you have an instinctive sense of what sounds natural. Some people are genuinely bilingual or near-native in more than one language and may translate into multiple target languages, and in some language pairs a shortage of native-speaker translators may mean non-native translation followed by native-speaker editing.
By contrast, there is no clear maximum number of source languages. One way to understand this is Roger Bell’s model of translation as two distinctive cognitive processes: decoding (comprehending the source text and converting it into an internal, non-verbal representation of meaning) and encoding (expressing that meaning clearly in the target language). A translator’s encoding skill stays largely the same regardless of how many source languages they add, so expanding into additional source languages mainly only requires comprehension of that language. In practice, translators also narrow what they learn to their specialism. For example, a legal translator may prioritise financial and business terminology over everyday topics, so they may “read” about a particular topic in a language well without feeling fully fluent speaking that language.
What makes a good translation? (1)
Assessing translation quality is complex and has been debated for centuries, with no single definitive answer; it depends on the lens you use. One major approach is functionalism, which judges a translation by whether it achieves its intended purpose for the target audience. For example, an instruction manual is “good” if it gives clear guidance, but conventions differ across languages: English often uses imperatives (“Press A”) in these sorts of documents, whereas German may prefer impersonal passives and French often uses infinitives. In sensitive genres such as patient consent forms, cultural norms, taboos, and expectations about how health information is discussed can strongly affect whether the translation has an equivalent effect. The same applies to marketing, where the goal (persuasion) is shared but the best way to achieve it may differ. Functionalist thinking therefore allows considerable freedom, even substantial rewriting in translation, as long as the end goal is met. In contrast, objective metric-based approaches compare translation and source line by line, categorising differences and scoring errors to provide a more reportable measure of quality, which can appeal in commercial settings. However, these systems can still be subjective and may overfocus on micro-level choices while missing overall effectiveness. In practice, many translators adopt a hybrid approach that balances detailed comparison with purpose and impact.
What makes a good translation? (2)
It is also possible to look at translation quality through a process-based lens rather than a product-based one. This is the approach taken in ISO 17100:2015 – the idea is that if a translation service provider follows a robust, well-defined workflow, it should produce consistently high-quality translations. There are two key criteria to highlight:
1) The four-eyes principle: Every translation should involve at least a translator (who drafts the translation) and a reviser (who checks it, primarily for accuracy). The ISO also recommends additional monolingual roles: a reviewer who focuses on style and a proofreader who checks for technical and grammatical issues.
2) The qualification criteria to follow when selecting translations. The ISO recommends three possible acceptable categories based on qualifications and experience:
translation qualification + 0 years of experience
other qualification + 2 years of experience
no qualification + 5 years of experience
Some agencies do, however, combine these and require a profession qualification plus two years’ experience. While understandable, this does limit opportunities for newly qualified translators and is not sustainable for the long-term future of the translation profession.
What makes a good translation? (3)
We’re ending this section on translation quality with a few key tips for how to use the ideas in practice:
Product-based
Be realistic about “quality scoring” models: They have a role in certain, limited contexts, but they aren’t the final word in translation quality.
Assess quality based on measurable content wherever possible: Pick materials with clear outcomes attached (e.g., website traffic, enquiries, conversions.
Use indirect evidence when outcomes aren’t easily measurable: gather signals from real users (e.g., customer complaints/questions, reader surveys, user feedback) to identify whether the translation is doing its job.
Process-based
Provide clear inputs at the start of the process: context, purpose, audience, reference material, previous translations, glossaries, style guides etc. Even file format choices can affect outcomes.
Close the loop with feedback: share performance information with your translator/agency (both positive and negative) so they can replicate successes and refine the process over time.
This will be more effective if you’re working with the same service provider rather than switching between providers for each project.
It’s not so much that one approach is better than the other. A really good translation quality model will draw on both product- and process-based understandings based on the actual real-world translation project.