Vocabulary Word List Sources: Where a Deck's Words Come From

If you want to know how big each CEFR level is, our counts are in how many words you need for A1, A2, B1 and B2. Here we look at a different question: not how many words, but whose words they are.

Why the source of a word list matters

Three things go wrong when a deck hides its sources.

First, you cannot tell what a level label means. “B1” on one app can mean something different on another, because each deck draws its levels from a list of its own. A deck’s level labels are only as meaningful as the list behind them.

Second, you cannot tell what you may rely on. A pronunciation, a meaning or an example sentence has an author or an owner, and a deck that names its sources lets you check them and tells you where to report a mistake.

Third, you cannot tell what was borrowed. Open data usually comes with conditions, such as an acknowledgement. A deck that lists its sources is also showing that it kept to them.

Cardday’s page puts this in a “Word data and sources” section, and the full notices are in the app under Settings, About. The table below gathers them in one place.

Levels: the CEFR-J Wordlist

The CEFR-J Wordlist is a graded list of English words compiled by the Tono Laboratory at the Tokyo University of Foreign Studies, under the leadership of Yukio Tono. It covers four levels: A1, A2, B1 and B2. The compilers say it was built from a textbook corpus, compared with the English Vocabulary Profile, and extended with words found only there.

Its terms are plain. The download page says it can be used for research and commercial purposes with a proper acknowledgement of the source, and it warns that the list may contain errors. That is why Cardday’s notice names the list, its version and its compiler.

A note on versions. Cardday’s notice names Version 1.5, while the download page now lists a later Version 1.6. We name the version we used because a list can change between releases, and a reader should know which one a deck follows.

The list counts entries by part of speech, so one headword can appear more than once with different levels. When you compare any deck with it, ask whether the count is words or entries.

It also helps to know what a list reflects. The compilers say the CEFR-J list was built from a corpus of English textbooks used in China, Korea and Taiwan, compared with the English Vocabulary Profile. That is a sound, documented basis, and it is also a particular one. No word list is neutral, so a named source lets you decide how far its levels fit your own learning. For that reason we describe the levels as a guide to order and difficulty, not as a promise about what you will be able to do.

Pronunciation: the CMU Pronouncing Dictionary

The Carnegie Mellon University Pronouncing Dictionary, usually shortened to CMUdict, is a free English pronouncing dictionary maintained by the Speech Group in Carnegie Mellon’s School of Computer Science and built for speech technology. Cardday’s notice credits it, with Carnegie Mellon’s copyright of 1993 to 2015, and that is why each card shows a US pronunciation.

Its terms are short. The maintainers state that use for any research or commercial purpose is unrestricted, they ask anyone who uses or redistributes it to acknowledge the origin, and they say they do not guarantee its accuracy and expect some errors and omissions. We take the last point as a reason to give readers a way to report a mistake, which we come back to below.

One limit is worth saying plainly. The dictionary gives written pronunciations; it does not record voices. When you tap the speaker on a card, the app hands the English word to your device’s own voice, which speaks it, so the sound you hear depends on your phone.

Order inside a level: frequency counts

Inside one level the words are not equal. A learner meets some words daily and others rarely. Cardday orders the words inside a level by how often they appear in everyday English sentences, using frequency counts derived from Tatoeba.

Tatoeba describes itself as a collection of sentences and translations, built by contributors. Its front page says its data is released under various Creative Commons licences, and Cardday’s notice states the licence that applies to the data it used. The notice also says that no Tatoeba sentence is included: the counts set the order, and the app shows its own sentences.

Frequency order is a design choice, not a rule. It puts the words you are most likely to meet again earlier in each level, which is useful when you are learning a few words a day.

What we wrote ourselves

Open sources supply the skeleton. The parts a learner reads most were written for the app: the Turkish meanings, and the example sentences with their Turkish translations.

We take that seriously because a meaning is a judgement. Some words have a second meaning, which Cardday’s page mentions. Sense notes from Open English WordNet 2025, which is based on Princeton WordNet and released under CC BY 4.0, helped while writing the cards. Open English WordNet describes itself as a free, open-source lexical database of English that groups words into sets of synonyms, each with a short definition. None of its text is in the app.

If you find a wrong meaning, Cardday’s support page says how to report it: open Settings, then About Cardday, tap Support, and include the English word and the app version shown in About.

How to judge any vocabulary app’s deck

Five questions, in the order we would ask them.

1. Where do the words come from? A named list is better than “carefully selected”.
2. What do the levels mean? Look for a named framework and a named list, not only labels.
3. Where do the pronunciations come from? And is the sound recorded, or spoken by the device?
4. What was written for the app, and what was borrowed? A deck that says so can be checked.
5. How do you report a mistake? A deck that cannot be corrected will not improve.

None of these tells you whether a deck is good for you, but they tell you whether it can be trusted. For how the words are then scheduled, see our explanation of how FSRS schedules reviews. On Cardday’s own page, the Cardday listing shows what the deck covers: English words with Turkish meanings, from A1 to B2. Its widget idea is the subject of our note on designing home screen widgets.