Dictionary or context? When to look up words while reading

Learners get two opposite pieces of advice about unknown words. One says: never stop, guess from context, let the language wash over you. The other says: look up every word, or you’ll never learn them. Both are partly right, and followed strictly, both will slow you down.

Here is a sentence that shows the problem:

Al final, quedamos en vernos el jueves.

Suppose you know every word except quedamos. Context helps a little: something about seeing each other on Thursday. So you look up quedar, and the dictionary’s first meanings are “to remain” and “to stay.” In the end, we stayed in seeing each other on Thursday? The sentence means “In the end, we agreed to meet on Thursday.” The meaning you needed was further down the entry.

This article is about when to look a word up, when to keep reading, and why the meaning in this sentence matters more than the dictionary entry.

What guessing from context can do

Guessing from context has a strong reputation, and for good reason. Native speakers learn much of their vocabulary from reading rather than from being taught it. But the learning from any single encounter is small. In a classic study, Nagy, Herman and Anderson (1985, “Learning words from context”) estimated that eighth graders reading a textbook passage in their own language had roughly a 5 to 15 percent chance of learning an unfamiliar word from a single encounter, depending on how strictly learning was measured. A later review of twenty experiments put the figure at about 15% of unknown words (Swanborn & de Glopper 1999, “Incidental word learning while reading: A meta-analysis”). Vocabulary grows through reading because of volume, not because each encounter teaches much.

For second-language readers, there’s another catch: guesses are often wrong. Bensoussan and Laufer (1984, “Lexical guessing in context in EFL reading comprehension”) asked first-year university students to translate a set of words, first on their own and then, a week later, with the text in front of them. Context helped in only about 13% of their responses, and their guesses came more from preconceived ideas about the words than from the context. More proficient students knew more words, but they weren’t better at using context.

Context works best under two conditions.

You understand almost everything around the word. This is the core of the 98% rule: when nearly all the words are familiar, the few unknown ones are surrounded by enough meaning to infer from. When one word in ten is new, context collapses.

The context actually points to the meaning. Compare two sentences, each with one word you may not know:

El perro del vecino ladra toda la noche, y nadie en la calle puede dormir.

A dog, all night, nobody can sleep: ladra almost has to mean “barks.” Now try:

Ella tomó la cartera y salió.

She took the cartera and left. Was it a wallet, a handbag, a briefcase or a school bag? Context can’t tell you. (Dictionaries list several of these, and usage varies between Spanish-speaking countries.) Most of the time, it doesn’t matter for the story. Sometimes it does.

What looking words up can do

Looking words up has a better research record than its reputation suggests, with an important twist. Hulstijn, Hollander and Greidanus (1996, “Incidental vocabulary learning by advanced foreign language students: The influence of marginal glosses, dictionary use, and reoccurrence of unknown words”) had Dutch university students read a French short story in one of three ways: with glosses in the margin, with a dictionary, or with neither. The students with glosses remembered the most target words. The students with a dictionary did no better than those with nothing, for a simple reason: they hardly used it, looking up only about two of the sixteen target words each. But the words they did look up, they remembered well. Looking up costs effort, and readers ration it. Glosses take the effort away. The same study found that words met three times were learned better than words met once.

Laufer and Hulstijn (2001, “Incidental vocabulary acquisition in a second language: The construct of task-induced involvement”) proposed one reason lookups help: a word you needed, searched for, and then fitted to its sentence gets more mental processing than a word you skimmed past. Their idea, now usually called the involvement load hypothesis, suggests that need, search and evaluation each make a word more likely to stick.

Glosses, short explanations attached to the text, have been studied a great deal, especially in digital reading. A meta-analysis by Abraham (2008, “Computer-mediated glosses in second language reading comprehension and vocabulary learning: A meta-analysis”) pooled eleven studies and found that computer glosses had a medium-sized effect on reading comprehension and a large effect on incidental vocabulary learning.

So looking up isn’t a failure of nerve. The costs are real, though. Every lookup interrupts the story. Too many, and reading turns into decoding, which is slow, tiring and easy to give up.

The dictionary problem: one word, many meanings

The deeper trouble with a dictionary is that it answers a different question. You want to know what a word means here. A dictionary tells you everything it can mean, anywhere. For many common words, that’s a long list.

Take Spanish quedar:

• Quedamos a las ocho. We agreed to meet at eight.
• Esa chaqueta te queda bien. That jacket suits you.
• No queda leche. There’s no milk left.
• Se quedó dormido en el tren. He fell asleep on the train.

Or German aufheben:

• Er hob den Brief vom Boden auf. He picked the letter up off the floor.
• Heb das Stück Kuchen für morgen auf. Save that piece of cake for tomorrow.
• Das Verbot wurde aufgehoben. The ban was lifted.

Or French prendre:

• Je prends un café. I’ll have a coffee.
• Ça prend du temps. It takes time.
• Elle a pris froid. She caught a cold.
• Il s’en est pris à son frère. He took it out on his brother.

The most common words are the most slippery, precisely because they’re used in so many ways. And some words can’t be understood alone at all, because they belong to a phrase whose meaning isn’t the sum of its parts:

Phrase	Word by word	Meaning
tomar el pelo (Spanish)	to take the hair	to pull someone’s leg
poser un lapin (French)	to put down a rabbit	to stand someone up
Tomaten auf den Augen haben (German)	to have tomatoes on your eyes	to miss something obvious
in bocca al lupo (Italian)	in the mouth of the wolf	good luck

Look up each word in Me estás tomando el pelo and you’ll learn a lot about hair. You won’t learn that someone is teasing you.

There’s a grammar version of the same problem. In Spanish, fuimos can come from ir (we went) or from ser (we were). Fuimos al mercado is “we went to the market”; fuimos amigos is “we were friends.” A word list can’t choose for you. The sentence can.

A simple rule for when to look up

When you meet an unknown word, ask three questions:

1. Does the sentence still make sense without it? If you can follow what’s happening, keep reading. You’ll probably meet the word again, and the next context may make it clear.
2. Does the story depend on it? If the word is the reason a character is afraid, or the object everyone is looking for, look it up.
3. Have I seen it before? The second or third time a word stops you, it’s earned a lookup. Words that keep appearing are the ones worth learning.

Add a fourth, unofficial question: am I curious? Reading in a new language should be a pleasure. If a word intrigues you, look it up.

If you find yourself answering “look it up” several times on every page, the problem isn’t your method. The text is too hard. Drop a level for a while; see how to choose your reading level.

How to look up well

Once you’ve decided to check a word, a few habits make the lookup count:

• Read the meaning for this sentence, not the whole entry.
• Note the base form. If fuimos stopped you, it helps to know it belongs to ir (or ser), so you recognize fue, iban and vamos later.
• Check the phrase. If the word sits inside an expression, learn the expression.
• Reread the sentence. The point of looking up is to understand the sentence, so finish the job.
• Hear it. A word you can say is easier to remember than one you only recognize on the page.
• Save it only if it’s worth it. A word you’ve met three times, yes. A word for a kind of 18th-century sail, probably not.

When to translate the whole sentence

Sometimes you know every word and still can’t follow. That’s a grammar problem, and looking up words won’t fix it.

Se me olvidaron las llaves.

Every word is easy. The structure isn’t: literally, something like “the keys forgot themselves to me.” It means “I forgot my keys.” Or in German:

Den Brief, den du mir geschickt hast, habe ich nie bekommen.

Here the object comes first and the main verb comes last: “I never received the letter you sent me.” When a sentence’s shape is the obstacle, a translation of the whole sentence is the right tool. Read it, then go back and read the original again, mapping each part. The second reading is where the learning happens.

Monolingual or bilingual?

Teachers sometimes insist on dictionaries written entirely in the target language. The research doesn’t clearly support that rule. Yoshii (2006, “L1 and L2 glosses: Their effects on incidental vocabulary learning”) compared glosses in Japanese learners’ own language with glosses in English, the language they were learning, and found no significant difference in how many words they learned. In practice, an explanation in your own language is faster and clearer up to about B1. From B2, explanations in the new language become more useful, and reading them is more practice.

How Lexo helps

Lexo’s word card is designed around this distinction between a word’s dictionary entry and its meaning in context. Tap a word, and the card shows:

• what the word means in this sentence, not a list of every sense;
• its base form and part of speech, so fuimos points you to the right verb;
• the meaning of the phrase or idiom it belongs to, so tomar el pelo is explained as a whole;
• a translation of the whole sentence into your own language, for the times when grammar is the problem.

You can play any word on its own to hear it pronounced, and save words to your Words list. For the hard paragraphs, a sentence-by-sentence mode shows one sentence at a time with its audio and translation.

To be straightforward about how this works: Lexo’s glosses and translations, like its adapted texts, are produced with AI models. Like any reference, they can occasionally be wrong, and if you spot a mistake, tell us.

Lexo is an iPhone and iPad app with about 100 classic novels at CEFR levels A2 to C2, in about 80 languages. It’s in beta and launching on the App Store soon. Get notified at launch.