Words or Sentences? Pick by Which Failure You Actually Have

Almost every article on this tells you to learn phrases rather than isolated words. That is a real answer, but it is an answer to only one of the two problems people have. Word-first study fixes coverage. Sentence-first study fixes retrieval. Diagnose which one is failing before you choose the unit.

The question is usually asked backwards

“Should I learn words or sentences” gets treated as a question about method, as though one is simply more effective. It is really a question about what you are currently unable to do, and the two units target different failures.

If you open an article and stop four times a paragraph because you do not recognise a word, no amount of phrase study will fix that quickly. You need volume, and volume is cheaper per word when the unit is small.

If you read comfortably but freeze when you have to produce a sentence, more words will not help either. You already have the words. What you lack is the machinery around them: which preposition follows, which noun the adjective usually modifies, whether the word is even the sort of thing you say out loud.

Two failures, two units. Picking by temperament instead of by symptom is why so many people study for a year and improve at the thing they were already good at.

What word-first actually buys you

Speed of coverage, and nothing else. That is not a criticism; coverage is often the binding constraint.

A bare word with its meaning takes a few seconds to review. A sentence takes several times that, because you read the whole thing, and in a review session of fixed length that difference compounds into hundreds of words a month. If your exam has a reading section and you are losing marks to unknown vocabulary, that arithmetic is the whole game.

The second thing word-first buys, less obviously, is coverage of the words you would never meet in curated example sentences. Textbook phrase collections skew toward useful, frequent, teachable language. The long tail of academic vocabulary does not live there.

What sentence-first actually buys you

Two things that a bare word cannot carry.

The first is collocation, the company a word habitually keeps. You can know that “consequence” means a result and still write “make a consequence” rather than “have consequences.” No definition contains that information. A sentence does, silently, and it does so without you having to study it as a separate fact.

Collocation is also the part of vocabulary that resists reasoning. There is no rule that explains why you make a decision but take a photograph, why an argument is strong but evidence is compelling, why you pay attention and give consideration. Native speakers are not applying a principle; they are recalling what they have heard. That is why the only efficient way in is exposure to the pairing rather than study of the pieces.

The second is register, the social level of the word. Bare definitions flatten everything to the same tone, which is how learners end up writing “the data was fantastic” in a paper or telling a friend the sandwich was “satisfactory.” An example sentence carries the setting along with the meaning.

There is a third benefit that gets oversold: fluency. Memorising sentences does make you faster at producing those sentences. It does less than promised for producing new ones, because a stored sentence is a unit, not a rule.

Diagnose yourself in two minutes

Take a page of something at your target level. A news article, an exam passage, whatever you actually need to read.

1. Read it once and count the words you could not understand at all.
2. Then take five words from that page that you did know, and write one sentence with each, without looking anything up.
3. Compare. If step 1 produced a long list, you have a coverage problem. If step 1 was fine but step 2 produced sentences you are unsure about, you have a retrieval problem.

Most people find step 2 harder than they expected. That is the useful part of the test: recognition feels like knowledge right up until you have to produce something. The distinction is worth understanding properly, and active versus passive vocabulary covers it in its own right.

What is going wrong	Study unit	What to actually do
Too many unknown words when reading	Bare word	High volume, short reviews, accept shallow knowledge for now
Know the word, use the wrong preposition	Sentence	Keep the example you met it in, not a constructed one
Writing sounds oddly formal or oddly casual	Sentence	Note the setting alongside the meaning
Can read fine, freeze when speaking	Phrase	Drill a small set you will genuinely reuse
Understand while listening, cannot say it	Word plus audio	Hear it every review, not just read it
Words vanish a week after learning them	Either	The unit is not the problem. The schedule is

That last row matters more than the rest of the table. If your words disappear regardless of how you stored them, changing the unit will not save them and you are solving the wrong problem. That is a spacing question, and spaced repetition is the thing to fix first.

The answer changes as you get better

The right unit is not a fixed property of you. It moves as your level does, which is why advice that worked two years ago can quietly stop working.

Early on, coverage dominates everything. When a page has fifteen unknown words, nothing else you do matters much, and the fastest route through is the cheapest possible unit. Sentence study at this stage is a luxury purchase: pleasant, slow, and aimed at a problem you do not yet have.

In the middle, the two costs are roughly balanced, and this is where most learners get stuck for years. Reading is mostly comfortable, writing is mostly stiff. It is also where the temptation to keep doing what already works is strongest, because adding more words still feels productive and still produces a visible number.

At the top, coverage nearly stops mattering. An advanced reader meets maybe one unknown word a page, and that word is usually inferable. What separates a C1 writer from a C2 one is almost entirely collocation and register, which means the unit has to be the sentence by then whether you enjoy it or not.

The practical implication: reassess every few months rather than picking a method and defending it. The person who needed bare words in March may need phrases by September.

The cost nobody mentions

Each approach has a failure mode its advocates skip.

Word-first produces confident wrongness. You learn that meticulous means careful, and then you write “a meticulous storm.” The definition was right; the range was invisible. You will not notice this on your own, because nothing about a bare definition signals what it does not cover.

Sentence-first produces volume collapse. Sentences take longer to review, so people quietly reduce how many they add, and a year later they have four hundred well-understood phrases and still cannot read a newspaper. The method feels better while producing less.

Neither failure announces itself. Both look like progress from the inside, which is why picking by symptom rather than by preference matters so much.

Where the two actually meet

The practical answer, for most people most of the time, is a word as the unit of review with one sentence attached to it. Not a sentence as the unit. The distinction is small and it decides how long reviews take.

When the word is the unit, you can review it in seconds and glance at the example only when something feels uncertain. You get the coverage rate of word-first and the collocation information of sentence-first, and you pay for the sentence only when you need it.

This is also the arrangement that survives contact with a busy week. A method that requires ten minutes of careful reading per session gets abandoned; a method that works in ninety seconds on a train does not.

Two details decide whether this actually works in practice. The first is where the sentence comes from. An example you met in the wild carries the memory of the context; a sentence written to demonstrate the word carries nothing, and you can feel the difference on review, because the constructed one reads like a definition wearing a costume. The second is that you resist adding a second and third example. Every extra line lengthens the review, and lengthening the review is exactly how a sustainable habit turns into one you skip.

If you want a supply of real sentences rather than manufactured ones, the cheapest source is whatever you already read. Harvesting one line from an article you were reading anyway costs nothing, and it beats any sentence written to order.

The week before an exam

Everything above assumes ordinary study. Close to a test, the advice inverts: stop balancing and go entirely to your weak side.

• Reading-heavy exam, weak coverage: bare words only, no new sentences, accept that you are buying recognition rather than mastery.
• Writing or speaking test: stop adding new words altogether. Take twenty phrases you know you will need and make them automatic.

The mistake is doing both in the final week. Split attention in the last seven days generally produces two half-finished efforts rather than one finished one.

The short version

There is no universally correct unit. There is a correct unit for the failure you have right now, and that failure changes over time, so the answer changes with it. Test yourself before you pick, and test again in a few months.

For what it is worth, this is why Vocaby is built word-first with an example, IPA and audio attached to each of its 29,000+ entries rather than as a sentence deck: the word stays cheap to review, and the context is there when you need it. FSRS spaced repetition handles the bottom row of that table, which is the one most people are actually failing. You can browse the entries on the Vocaby word pages.

Test which failure is yours, then pick. Get Vocaby on the App Store