Translations or English-Only Definitions? What Research Says

Use both, at different stages. Research comparing first-language glosses with English-only definitions has generally favoured the translation for getting a word into memory, especially below advanced level. The familiar advice against translating is really about producing language later, not about how you first learn what a word means.

The advice everyone gives

Somewhere in your first year of study, someone told you to stop translating. Think in English. Use an English-English dictionary. Put the bilingual one away.

The instinct behind it is sound, and anyone who has watched a learner assemble a sentence by translating word by word knows what it’s guarding against. That process is slow, it produces sentences no native speaker would write, and it keeps your first language sitting in the middle of every exchange.

The trouble is that the rule gets applied to a different moment than the one it was designed for. There’s a difference between using English without translating and learning what a word means. The advice is good for the first and questionable for the second, and almost nobody separates them.

What the research actually found

The comparison has been studied directly, usually by giving learners a reading text with glosses in either their first language or in English, then testing what stuck.

The general pattern favours the translation. Work including Laufer and Shmueli (1997) and Miyasako (2002) found first-language glosses produced better retention than English-only ones, in both short-term and delayed testing. Later research has kept refining the picture rather than overturning it.

It’s not unanimous, and the details matter. Effects vary with proficiency level, with how the gloss is presented, and with how learning is measured — the same glossing method can look better or worse depending on whether you test recognition or recall. The clearest finding is that help in the first language is most useful at lower proficiency, where an English definition may cost more effort than the word being defined is worth.

That last point deserves unpacking, because it’s where a lot of study advice quietly misleads. A method can look excellent on a multiple-choice test and do very little for you in conversation, since recognising the right answer among four options is a far easier task than producing the word from nothing. When you read that some technique improved vocabulary scores, the question worth asking is which kind of score, and the answer is usually recognition, because recognition is cheap to test at scale.

So the honest summary isn’t “translation wins.” It’s that translation is at least as good as English-only for acquisition, better for many learners, and the blanket advice against it isn’t supported by the evidence.

Why translation works: it’s cheap

The mechanism is straightforward once you think about what each option costs.

Your working memory is limited, and everything you do while learning a word competes for it. An English-only definition asks you to parse a sentence of English, possibly containing words you also don’t know, and infer a meaning from it. That’s real cognitive work, and it happens before any learning of the target word begins.

A translation gets the meaning across almost instantly, at close to zero cost. Whatever capacity you have left goes into encoding: noticing the form, the sound, the example. When the definition is expensive, the word gets less attention, not more.

There’s a second effect too. A word you understand immediately feels learnable, and a definition you had to fight through feels like a chore. Over hundreds of words, that difference in friction decides whether you keep going.

Where translation goes wrong

None of this means bilingual study is harmless. It has one specific failure mode, and it’s serious enough to explain why teachers warn against it.

The failure is storing a rigid one-to-one pair. If a word is filed in your head as the single equivalent of a word in your language, you’ll reach for it whenever that word appears in your thoughts, including all the situations where English does something else. This is the machinery behind sentences that are grammatically perfect and unmistakably foreign.

Real vocabulary doesn’t map one-to-one. A single word in your language may split into three English words with different registers, and one English word may cover ground that your language divides up. A pair can’t hold that. An example sentence can.

That’s the actual lesson: the problem was never the translation, it was the format. A bare pair is a bad way to store a word regardless of what language the right-hand side is in.

The same word, stored two ways

It’s easier to see with an example. Take a learner who files this entry:

appointment = a meeting arranged in advance

Fine as far as it goes. Now watch what happens when they want to say they’re busy on Friday evening. Their language likely uses one word for the dentist, the client meeting, and drinks with a friend, and the entry they stored maps that word to appointment. Out comes “I have an appointment with my friend,” which is grammatical, comprehensible, and wrong in a way that’s hard to explain.

Now the same word stored as a chunk:

appointment — make an appointment, cancel an appointment, a doctor’s appointment, a 3pm appointment “I’ve got a dentist appointment on Friday, so I can’t make lunch.”

Nothing about the second version forbids using a translation to grasp the meaning. What it adds is the company the word keeps, and that company is what tells you appointment belongs to dentists and clients rather than friends. The learner with the second entry never generates the bad sentence, because the pattern they stored doesn’t produce it.

That’s the whole difference, and it costs about fifteen extra seconds per word.

What English-only definitions are genuinely good for

They have real strengths, and they arrive later than most people are told.

Reading a definition in English exposes you to more English, and the definition itself is often built from useful high-frequency vocabulary. It also forces you to think about the concept in the target language rather than reaching for an equivalent, which builds the network of associations that makes recall faster once the word is already known.

Most importantly, English-only definitions capture nuance that translations flatten. Two words that share a translation in your language often differ in formality, in strength, or in what they collocate with, and only an English source will tell you that. This is the same territory as English collocations, where the useful information is about which words appear together rather than what any single word means.

The catch is that all of this requires enough English to read the definition comfortably. Below that threshold you’re not gaining nuance, you’re just paying more for the same meaning.

Which to use, and when

Situation	Better choice
First encounter with a word, below advanced level	Translation
First encounter, advanced level	Either, English-only if comfortable
Concrete nouns and everyday verbs	Translation, it’s exact and fast
Abstract or culturally loaded words	English definition plus example
Distinguishing two words that share one translation	English-only, definitely
Checking register or formality	English source
Reviewing a word you already know	Recall from meaning, no gloss at all
Getting through a long text without stalling	Translation, speed matters more here

The pattern: use your first language to get in the door, and English to understand what’s inside.

The hybrid that avoids both traps

Practically, this comes down to a few habits.

• Take the translation, then don’t stop there. Get the meaning quickly, then attach an example sentence in English. The example is what prevents the one-to-one pair from forming.
• Store chunks, not equations. Alongside the meaning, keep two or three words the target word usually appears with.
• Add the sound immediately. A word learned only through text stays silent, and silent words fail in conversation whichever language you defined them in.
• Switch to English-only for near-synonyms. Whenever two words come out identically in your language, a bilingual entry cannot help you, by definition.
• Drop both when reviewing. Recall the word from its meaning without looking, which is where the actual strengthening happens. That mechanism is covered in active vs passive vocabulary.
• Stop translating when you produce. This is where the original advice is right, and it works because your vocabulary is solid enough not to need the crutch, not because you refused the crutch while learning. We wrote about that shift in how to think in English.

How Vocaby handles it

Vocaby is built around the format that makes this work rather than around picking a side. With more than 29,000 words, each entry carries the meaning together with example sentences, so the word arrives as a usable chunk rather than a pair to memorise.

Every word also carries IPA and real audio. That matters here more than it first appears: the one thing neither a translation nor an English definition gives you is the sound, and a word you can define in two languages but not recognise when spoken is still missing the half you need for conversation.

Then FSRS spaced repetition handles the reviewing, which is the stage where both gloss types stop mattering and retrieval takes over. It brings each word back just before you’d forget it, so the meaning you got cheaply on day one is still there in week six. Curated decks let you focus, and swiping keeps the daily habit small. You can browse the range in the word library.

Take the translation. Don’t file it as a pair. That’s the whole method.

Download Vocaby on the App Store to learn words with meanings, examples, audio, and spaced review on every entry.