Should you learn words or whole sentences?

Learn the word with its meaning first. Then add sentences to it — more than one, because one example sentence changes nothing. Vary the kind of sentence later, once the word is no longer new.

Do it in that order and more of the words are still there days later. Two studies gave the same learners the same words, with and without a sentence around them. The words that came with sentences came back better. In one of the two, learners also used those words better in sentences of their own.

Two habits sound like the same advice and cost you words instead.

Don't guess the meaning from the sentence. Look the word up first. Learners who guess end up knowing fewer words than learners who are told.

Don't expect the sentence to be free. Those extra sentences are extra meetings with the word. That may be where the gain comes from.

One thing decides how much any of this is worth to you. Recognising a word and coming up with it are different jobs. The order above makes a word easier to recognise, and easier to pick when a sentence needs it. Getting it out of your mouth unprompted is a separate problem, and most of what the intermediate plateau is made of.

Does adding an example sentence help you remember a word?

Yes, if the sentence comes on top of the meaning instead of in place of it. And if there is more than one sentence.

One sentence does nothing. Webb tested a plain word pair against the same pair with one example sentence. He measured it ten ways, and the two came out level on all ten.

Three sentences helped a week later. 33 learners got twenty hard English words with their translations. Half also got three sentences. Those learners wrote better meanings, and better sentences of their own.

Ten meetings helped a lot. 72 adults studied Lithuanian word pairs, then read a story. Two-thirds of the words appeared in it, ten times each. Two days later the learners remembered 62% of the story words and 35% of the rest.

Some of that gain may be nothing more than meeting the word more often. The authors of one study say they cannot tell the two apart.

Your own sentences are a different matter. Timing decides them. Barcroft's learners of Spanish wrote brand-new words into sentences. They then produced fewer of those words than learners who studied each word beside a picture. The writers had twice as long with each word and still lost.

So write your own sentence once a word has a foothold, not at first meeting.

How many different sentences should a word turn up in?

Varied sentences don't make a word stick harder. They make it portable.

You can know the word for "bill" as the thing you ask for at the end of a meal. The portable version of the same word also covers your phone bill and your bank account.

239 adults learned invented words from ten sentences each — ten on one topic, or ten on ten topics. One topic made a word easier to pick out for a familiar kind of sentence. Ten topics made it easier to pick out for a sentence the learners had never seen. Neither helped with coming up with the word itself.

The foreign-language version held the meetings equal, with 124 learners. The single-topic group was ahead on the day. The varied group was ahead a week later. That win was on recognising the meaning rather than recalling it, and it was worth a word or two in ten.

Variety was worst for the learners with the smallest vocabularies. So don't vary a word you met this morning. Anchor it in one context, then diversify. Spreading a brand-new word across five stories left learners remembering less of its meaning, in a pre-registered study of 100 adults.

Nobody has checked whether varied sentences help you say a word unprompted.

Is sentence mining effective?

Nobody has tested sentence mining. No published study compares sentence cards against word cards in a spaced-repetition system.

What has been tested is the shape of the card. A mined card is a line from a show with the word blanked out. So the card built to escape fill-in-the-blanks is a fill-in-the-blank.

That shape won the biggest comparison anyone has run. Across 154 students, three fill-in-the-blanks beat one self-written sentence. What mattered was the number of goes, not how hard the task was. The goes weren't matched either, which was the author's own point.

So mine sentences if that is how you enjoy collecting material. Don't expect the card to do something a blanked-out line can't.

Should I learn phrases instead of single words?

Collect phrases, but not as a shortcut. They are heavier than single words.

Researchers ran single words and two-word phrases through the same activities. The phrases were clearly harder to produce.

The case for them is on the other side. 32 students worked through 22 hours of listening and reading. Half of them had set phrases pointed out along the way. Judges who didn't know the groups rated that half as sounding more proficient in an interview.

So "learn phrases, not words" is a good argument about what is worth having. It is a bad one about what is easy to get.

What to do, in order

1. Take the word and its meaning first, on its own. Write the meaning in your own language rather than the one you are learning, which beat doing it the other way round.

2. Then add sentences. More than one of them. If your deck gives one example sentence per word, that sentence is decoration.

3. Vary the sentences once the word isn't new. Change the subject, not just the wording.

4. Count your goes before you count your effort. Three quick passes at a word beat one elaborate pass in the biggest test of it. The goes weren't matched, so hold it loosely.

5. Write your own sentence once the word has a foothold. Minute for minute, reading three sentences containing a word beat writing your own with it. Writing won when it got the time it needs — "if the second experiment represents authentic learning", as Webb hedges it.

None of this needs an app. All five are a deck, a notebook, and the order you use them in.

Where Pezmu fits

Flashcards do the word and its meaning well. They do the sentences around it badly, and the field's own flashcard specialist says so.

Tatsuya Nakata studies flashcards for a living, and publishes a vocabulary app of his own. His verdict on cards: "Flash card learning is not very effective in promoting generative use because target items are usually encountered without context or in a single context." Generative use means meeting a word you already partly know in a new context. His prescription is more reading, listening and watching. Ours is a conversation.

Pezmu is typed chat with a tutor that remembers you. Underneath it keeps a record of every word you have saved, and whether you have said it yet.

Tap a word in a reply and you are told what it means. Tap again to save it. That is the meaning first, on its own.

After that the tutor uses the word back at you when it fits what you are talking about. Not every turn, and only where it genuinely fits. That is the sentences around the word, without you building a card for them.

Then the only thing we count. A word goes from Saved to Hinted to On your own. It reaches that last stage only when you say it in a message with nothing in front of you showing it to you.