Learning synonyms in the same sitting makes each one harder to remember, not easier. Because near-synonyms share most of their meaning, your memory has almost nothing to hang the difference on, and the words blur together. In a 2022 study, learners recalled unrelated words 56% of the time against 43% for words taught in a meaning-based cluster.
That number should be annoying, because meaning-based clusters are how almost every vocabulary resource is built. Ten words for happy. Fifteen words for said. Twenty adjectives to describe personality. The format is so standard that nobody questions it, and it turns out to be close to the worst possible arrangement for the moment you are actually trying to get the words into memory.
The list that teaches you the least
Picture the list. Glad, cheerful, content, elated, jubilant, buoyant, merry, upbeat. Eight words, one page, one afternoon.
You get to the end feeling like you have learned eight words. A week later you can produce maybe two, and if someone asks you which of elated and jubilant is stronger, you have no idea. What you actually retained was a vague cloud that means roughly happy, with eight labels floating loose above it, unattached to anything.
That is not an effort problem. The only thing available to tell those eight apart is the thin slice of meaning where they differ, and you tried to take in eight of those slices in one go.
What happens in your head when two words overlap
New vocabulary gets stored with whatever hooks you can give it: sound, spelling, the sentence you met it in, a mental image, the emotion of the moment. Distinct words come with distinct hooks for free. Anchor and velvet have nothing in common, so nothing competes.
Near-synonyms arrive with the same hooks. Same rough meaning, same register, often the same part of speech, frequently the same context in the list. The only thing separating elated from jubilant is a shade of intensity and a difference in how they are used, which are exactly the two things a beginner does not have yet.
So the memory traces overlap, and when you reach for one you get a smear of all of them. Psychologists have called this interference since the 1930s, and it is not specific to language: it happens whenever you learn several highly similar things in a single window.
Two different things are going wrong here, and from the inside they feel like one thing. The first happens at storage: the words go in badly because there is nothing to store as the difference. The second happens at retrieval: the words went in fine, but when you reach for one, its neighbours come up as competing candidates and you cannot tell which is which. The synonym list gives you both at once, which is why it produces such a specific, recognisable feeling — you look at jubilant and know you have met it, you know roughly what neighbourhood it lives in, and you still cannot say it out loud in a sentence. Recognition survives. Production does not.
That gap is the useful signal. Recognising nine words from a list and being able to use two of them does not mean you saw the words too few times. It means nothing ever made them separate.
What the research found
Three studies, spread across twenty-five years, point the same way.
| Study | Setup | Result |
|---|---|---|
| Tinkham (1997), Second Language Research 13(2) | Compared meaning-based sets against situation-based sets and unrelated sets | Meaning-based grouping hindered learning; situation-based grouping helped |
| Erten & Tekin (2008), System | Presented new words in semantic sets versus unrelated sets | Words taught in semantic sets needed more repetitions before they were fully learned |
| Pérez-Serrano, Nogueroles-López & Duñabeitia (2022), Frontiers in Psychology 13 | 102 university students watched a subtitled video containing twelve target words | At eight exposures, recall was 56.4% for unrelated words against 42.6% for clustered ones; recognition 70.8% against 59.0% |
One detail in the 2022 study surprised me. The gap between the two conditions got wider with more exposures, not narrower. Both groups improved; the unrelated group just improved faster. So the intuitive fix, going back over the synonym list a few more times, is the one thing the data says will not close it.
Semantic sets and thematic sets are not the same thing
This is the distinction that makes the finding usable, and it is the one people miss.
A semantic set groups words that mean similar things. Eye, nose, ear, mouth, chin. Same category, same slot in a sentence, and to someone who does not know them yet, more or less interchangeable.
A thematic set groups words that belong to the same situation without meaning the same thing. Tinkham’s example is frog, green, hop, pond, slippery, croak. A noun, a colour, a verb, a place, an adjective, a sound. They hang together as a scene, so each one gives the others context instead of competing with them.
In Tinkham’s work thematic sets went further than avoiding the damage: they beat the unrelated control outright. So “learn words in random order” is the wrong lesson to draw. What matters is which kind of relatedness you are leaning on.
| Semantic set | Thematic set | |
|---|---|---|
| Example | strident, raucous, shrill, piercing | siren, dawn, wake, groggy, curtain |
| What they share | a definition | a situation |
| Parts of speech | usually identical | usually mixed |
| Effect on learning | interference | support |
Why every resource does it anyway
If meaning-based grouping is the weak arrangement, why is it also the universal one? The answer turns out to be mundane, and it changes what you think you are holding when you pick up a word list.
Categories are cheap to produce. “Twenty words for angry” writes itself: open a thesaurus, which is already indexed by meaning, and copy a column. A situation-based set has to be invented, then checked to see whether those words genuinely turn up together. One is a lookup. The other is writing. Volume has always followed the cheaper of the two.
Categories are also genuinely better for the jobs that are not learning. If you are writing and want a stronger word than angry, a meaning-based list is exactly the right tool, because you already know the concept and need the label. If you are revising words you already know, grouping them by meaning is a fast way to compare them. The format is good at what it was built for. It just gets handed to people at the one stage where it does the most damage, which is the stage most of us are in when we open it.
Where the evidence gets messy
Two honest caveats, because the finding is often repeated more confidently than the literature supports.
First, results are not unanimous. Some studies have found no meaningful difference between clustered and unclustered presentation, and a few report an advantage for clustering. The effect is real and has been replicated, but it is not a law of nature, and effect sizes vary a lot with the design.
Second, the effect is clearest at first encounter. It shows up when the words are genuinely new and are being encoded side by side. Once you already know cheerful solidly, adding jubilant next to it is a completely different task — now the word you know is a landmark rather than a competitor. Comparing synonyms is a great way to refine words you already have. It is a poor way to acquire words you do not.
That distinction resolves most of the apparent contradiction with how advanced learners study, and it is why “never look at synonyms” would be the wrong takeaway.
How to space similar words apart
The practical version is short, and I break the first rule constantly, so this is as much a note to myself.
- One per batch. When you add new words, let only one of any near-synonym group in. The rest go on a waiting list, not in the bin.
- Wait until the first one is solid. Solid means you can produce it in a sentence without hesitating, not that you recognised it once.
- Give each new word a different sentence, from a different place. The context is doing the work of separating them.
- Learn thematically when you want a set. If you want to study a topic, pull a scene rather than a category: mixed parts of speech, one situation.
- Use the non-meaning hooks. Sound and stress separate words that meaning does not. Elated and jubilant mean almost the same thing and sound nothing alike, and that difference is free storage.
If you want a sense of how many new words a batch should hold in the first place, we went through that separately in how many words to learn a day.
Concretely, here is what that looks like for the eight-word happiness list. Week one, you take elated only, and the other seven go on the waiting list. You meet it in a real sentence, you review it a few times, and by week three you can use it without stopping to think. At that point jubilant comes in — and now it has a job, which is to be the one that is louder and more public than elated. Week six, content arrives, and it defines itself by being the quiet one at the other end. Each word enters against a background that already exists.
Three words in six weeks sounds slow next to eight words in an afternoon. Six weeks after the afternoon, though, you have two of the eight. The slow version is ahead by then and does not lose ground afterwards, because each of the three has a distinct place rather than a shared cloud.
What to do with the synonym lists you already own
You do not have to throw them out. Change what they are for.
A synonym list is a bad learning sequence and a good reference. Keep it, and use it in the direction the research supports: when a word you already know well comes up for review, look at the list and ask which neighbour is a shade stronger, which is more formal, which one you would never use in speech. That is refinement on a foundation, and it is where comparing words pays off.
The other use is diagnostic. Go down a list of near-synonyms and mark the ones you could actually use in a sentence today. Usually it is one or two out of ten. That tells you the list never taught you the other eight, which is worth knowing before you spend another afternoon on it.
How Vocaby handles near-synonyms
Vocaby uses FSRS spaced repetition, which schedules every word on its own interval based on how you personally did on that word. That has a useful side effect here: two synonyms added on the same day stop travelling together almost immediately. The one you find easy slides out to a long interval while the harder one keeps coming back, so within a couple of weeks they are being reviewed on completely different days. You do not have to plan the separation; it falls out of the scheduling.
The other half is the hooks. Every one of the 29,000+ words in Vocaby carries its IPA transcription and audio, so a word arrives with a sound and a stress pattern attached and not only a definition. When two words mean nearly the same thing, how they sound is often the only thing that is genuinely different about them, and it is the part a written list leaves out entirely. You can see how an individual entry is put together on any of the word pages.
None of this makes a synonym list a good idea on the day you meet the words. It just means you are not fighting your own review schedule as well.
Learn one word properly, then meet its neighbours later, when there is something for them to be different from.
Get Vocaby on the App StoreFrequently asked questions
- Is it bad to learn synonyms at all?
- No. The problem is timing, not the words. Learning glad, cheerful, and elated in the same session forces you to encode three nearly identical meanings at once, which is when they blur. Learn one now, meet the others weeks later, and each has something to attach to.
- What should I group words by instead?
- Group by situation rather than by meaning. A set like boarding pass, delay, gate, overhead, and aisle shares a scene but no definitions, so the words support each other instead of competing. Researchers call this thematic clustering, and it tends to help rather than hurt.
- Does this apply to word pairs like affect and effect?
- Partly. Those are confused for spelling and sound rather than meaning, so the fix is different. But the timing rule still holds. Introducing both in the same sitting is harder than learning one well and adding the other once the first is solid.
Keep reading
Ask, Question, Interrogate: Why English Has Three Words for Everything
English keeps a short Old English word, a French one from after 1066, and a Latin one borrowed by scholars. Ask, question, interrogate. Here is where the three layers came from, how to tell them apart, and which one to actually use.
Translations or English-Only Definitions? What Research Says
Everyone tells you to stop translating and learn words in English. The studies mostly say the opposite, at least for getting a word into memory. Here's what the evidence actually shows and how to use both without picking up bad habits.
Which Direction Should Your Flashcards Face?
Every flashcard runs in a direction. Word to meaning trains reading; meaning to word trains speaking. The research says they are not interchangeable, and the one that helps you most changes as your level rises.