Why Some Words Are Harder Than Others

A word’s difficulty is not a fixed property of the word. It is the gap between the word and what you already have, which is why the same item can be nearly free for one learner and expensive for another. Applied linguists call this a word’s learning burden, and most of it is predictable before you start.

Difficulty is a relationship, not a property

Everyone has a deck where one card keeps coming back while its neighbours graduate. The natural reading is that you are doing something wrong, or that your memory is failing on that item for no reason. Usually neither is true. That word costs more than the ones around it, and there are only a handful of reasons why.

The useful idea here is learning burden: the amount of effort a word takes to learn, given what the learner already knows. The key phrase is the second half. Burden is not a rating you could print in a dictionary, because it changes with the person.

For an English speaker learning Spanish, responsabilidad is close to free. The form is guessable, the meaning is identical, and the grammar behaves as expected. For the same learner, ojalá is expensive: unfamiliar shape, no English relative, and a meaning that does not sit neatly on any single English word.

Neither word is harder in the abstract. The difference is entirely in what the learner brought with them.

The biggest discount is your first language

Whatever overlaps with your first language is cheap, and the overlap is often larger than people expect.

Where an L2 word has similar associations to its translation in your first language, the burden is light. You are not building a concept, you are attaching a second label to one you already own. That is a much smaller job than it feels like, and it is why cognates disappear into your vocabulary almost without study.

Korean learners of English get a particular version of this. Korean has absorbed a large stock of English loanwords, so a word like stress or schedule arrives with the form already familiar. The catch is that the borrowed sense sometimes drifted, so the form is free and the meaning is a trap. That is a different bill from an unfamiliar word, and it wants a different fix.

It is worth being precise about what the discount actually covers, because people over-claim it in both directions. A cognate hands you the form and usually the core sense. It does not hand you the register, the collocations, or the grammatical behaviour. French demander and English demand overlap in meaning but not in force, and a learner who treats the pair as free will write sentences that are technically correct and socially wrong.

So the honest version is that L1 overlap removes one or two components of the burden rather than all of them. That is still an enormous discount. It is just not the whole bill.

Concrete beats abstract, with a caveat worth knowing

The concreteness effect is real and well replicated: concrete words are learned faster and remembered better than abstract ones. The usual explanation is that abstract words lack sensory and motor detail, so there is less to attach them to.

Here is the caveat that most summaries drop. That research is largely about first-language acquisition, where the learner is building the concept and the word together. An adult learning a second language already has a concept for justice or resentment; they are only mapping a new form onto it. Researchers have pointed out that this makes the concreteness advantage work differently in a second language than in a first.

So I would not use it to predict which of your cards will fail. Use it as one input among several, and expect it to matter less than it does for a child.

Where it does bite in a second language is at the edges of a sense rather than at its core. You know what resentment is. What you may not know is how much of it counts as resentment rather than irritation, and abstract words tend to have blurrier edges than concrete ones. That is why the fix for an abstract word is two contrasting examples rather than one definition: the boundary is the part that is missing, and a definition does not draw boundaries, it draws centres.

Form costs, quietly

Some of a word’s cost is nothing to do with meaning.

• Length and syllable count. Longer words carry more form to hold
• Pronounceability. A word you cannot say has no phonological handle, so you are storing a picture of letters rather than a sound
• Orthographic regularity. When spelling and sound disagree, the word has effectively two forms to learn
• Part of speech. Studies of learning burden and decay have found part of speech and word length both matter, with nouns generally cheaper than other classes
• Morphological transparency. A word built from parts you recognise costs less than an opaque one of the same length

The pronunciation one is the most actionable and the most skipped. If you have only ever seen a word, you have half of it, and the missing half is the one that has to work in real time when someone says it to you.

Transparency is the other lever you can pull deliberately. Malevolent is a long word, which looks expensive, but if you already hold mal- and bene- from elsewhere it costs less than a short opaque word like gist. Length is a poor proxy for difficulty on its own; length without recognisable parts is the expensive combination. Our notes on word roots cover that half.

What this does not excuse

An honest limit, because this framing is easy to abuse.

Learning burden explains why a word is expensive. It does not mean an expensive word is unlearnable, and it is not a licence to declare every stubborn card intrinsically hard. Most stuck words in practice are not exotic items with high intrinsic burden. They are ordinary words that were met once, in a gloss, with no sentence and no sound.

That is a different diagnosis with a different fix, and I have watched myself reach for the flattering explanation first. If the word is common and you still cannot hold it, the cost is probably in how you met it rather than in the word.

Reading the bill

The practical move is to work out which component is expensive, then pay that one.

What is happening	Likely cost	What actually helps
You cannot say it, even silently	Form, no sound attached	Learn the pronunciation before anything else
You know the meaning but never produce it	Mapping, not storage	Write one sentence of your own
The meaning feels blurry no matter how often you see it	Abstract, no edges	Two contrasting examples, not one definition
It looks like nothing you know	No morphological anchor	Learn one relative alongside it
You know a similar word in your language and keep using it wrong	False friend	Study the difference, not the word

The last row is the one that catches confident learners, because the form is free and confidence is high, which is exactly the combination that stops you checking.

Notice that four of the five fixes are one-off. You pay them once and the word gets cheaper permanently, which is different from review, where you pay a little forever. That is the argument for spending real effort on a stuck word rather than grinding it: you are trying to change its price, not to out-repeat it.

The one that is not one-off is the false friend, because the wrong version is already installed and competing. That one needs the correction plus a later check, which is a different shape of work.

When the right move is to drop it

Worth saying plainly: some words are not worth their price.

A word with high burden and low frequency is a bad investment, and there is no rule requiring you to finish every card you started. If a word is rare in what you actually read, and it costs four times what its neighbours cost, letting it go is a decision rather than a defeat. You can meet it again later, when you have more to hang it on, and it will be cheaper then because your inventory is bigger.

This is the part of vocabulary study people feel guiltiest about and should not.

There is a rough test. Ask when you last met the word outside your own deck. If the honest answer is never, and the word is also expensive, you are maintaining an item that your actual reading is not asking for. Suspending it costs you nothing you were getting. If you do meet it later, that encounter is worth more than the twenty reviews you would have spent in the meantime, because it arrives with a context attached.

Card problems versus word problems

These two get confused constantly, and the distinction decides what you should change.

A card problem lives in the presentation: an ambiguous prompt, an example sentence containing a second unknown word, a bare gloss with no context. Those are editable, and our notes on words you keep forgetting go through them.

A word problem lives in the word’s relationship to you: no cognate, no recognisable parts, an abstract sense, a shape you cannot pronounce. You cannot edit those away. You can only add what is missing, or decide the word is not worth it yet.

Before you rebuild a card, it is worth asking which of the two you are actually looking at. Rebuilding a card cannot fix a missing anchor, and adding an anchor cannot fix a badly written prompt.

What an app can take off the bill

Software cannot lower a word’s intrinsic burden. What it can do is remove specific components of it before you meet the word.

Every entry in Vocaby’s 29,000-plus word list carries IPA and audio, which kills the pronounceability cost outright, and an example, which gives an abstract word at least one edge to hold. The FSRS scheduler then does the part that follows from all this: words that turn out to be expensive come back more often, without you having to decide which ones those were.

That last bit is the quiet advantage. You do not have to estimate a word’s burden correctly in advance. You just have to notice, when a card keeps returning, that the honest question is which part of it is costing you, and not whether you are bad at this.

Get Vocaby on the App Store