Why Word Lists Fail

Memorizing the 2,000 most common words rarely becomes vocabulary you can actually use. The research points to what works instead.

An editorial illustration of a long printed word list crumpled at the edge of a desk, next to an open notebook where a few words are copied with short context lines.

The unfinished word list

Many language learners keep one: a downloaded "1000 most common words" file, unit 12 of a textbook, or a deck built from someone else's list. The first twenty entries feel like progress. Midway through, the remaining items get hard to tell apart. By the end, the file is still there, but it is no longer in use.

Plenty of words can be learned from a list. What fails is a particular habit: a generic list, studied once, never reviewed, with no connection to anything you read or hear. The reasons are well documented.

A list you never reopen follows the forgetting curve

The forgetting curve is the oldest finding in memory research. Ebbinghaus measured how much faster relearning was after different delays, and a modern replication matched the classic curve (Murre & Dros, 2015): savings fell to about 58 percent after 20 minutes, 44 percent after one hour, and 34 percent after one day, with the steepest drop in the first hours. Coming back to relearn is much cheaper than starting over, but only if you reopen the list.

A word list that is studied once and never reopened follows exactly this curve. Those numbers describe savings, not the claim that "you forget 70 percent of vocabulary in a day". Without a revisit, the savings keep falling after that first day, and the words drop out of reach.

A finished list does not give reading coverage

Hu and Nation (2000) rewrote fiction with 0, 5, 10, or 20 percent unknown words; most readers needed around 98 percent coverage of running words for adequate unassisted comprehension, and almost nobody managed at 80 or 90 percent (Hu & Nation, 2000). Later work splits the threshold: around 95 percent coverage as a minimum, around 98 percent as the comfortable zone (Laufer & Ravenhorst-Kalovski, 2010). Nation (2006) estimates that 98 percent of novels and newspapers needs roughly 8,000 to 9,000 word families (Nation, 2006).

The 2,000 most common words are worth learning, but that list does not by itself put those words into the texts you actually meet. If a page in front of you is ten percent unknown, coverage is 90 percent. That is the condition where Hu and Nation found that almost nobody managed adequate unassisted comprehension.

Retrieval sticks better than rereading

There is a reliable finding under the name test-enhanced learning: trying to recall something makes it stick better than studying it again, even when the extra study feels more productive in the moment (Roediger & Karpicke, 2006). A follow-up showed that repeated retrieval was the critical part of learning that lasts (Karpicke & Roediger, 2008).

Reading a list top to bottom, or highlighting it, is restudy, not retrieval. A list supports learning when you use it as a prompt: cover the word, try to say it, then check. The same retrieval step is the advice in our article on language exchange notes: catch a few words and reuse them next time.

Spacing beats cramming

Cramming a 500-word list over one weekend is massed practice. A meta-analysis of 317 experiments found distributed practice consistently beat massing, and the useful gap grows with how long you want to remember (Cepeda et al., 2006). Independent reviews rate both practice testing and distributed practice as high-utility techniques (Dunlosky et al., 2013).

Meeting fifteen of those words again next week, and fifteen more the month after, beats one long weekend with all five hundred. A printed chapter list rarely schedules those later meetings; a review system does.

The theme pack can work against you

Studying a whole semantic set at once can interfere with learning it. New words presented as a category (all fruits, all furniture, all colours) took more trials to learn than unrelated or story-linked sets (Tinkham, 199390027-E); 1997), and category presentation slowed both encoding and later translation of new labels (Finkbeiner & Nicol, 2003). Later classroom studies are more mixed, so this is not settled law. The practical advice from the same line of research: do not use a brand-new colour set or animal set as the first time you study those words.

Words that showed up in one real conversation are closer to a thematic set, which Tinkham (1997) found easier to learn than a category set. That is one reason conversation-caught words behave better than unit 12.

Lists can teach form and meaning fast

Compared with learning from context alone, translation-list study is often faster for remembering form and meaning (Prince, 1996), and word-focused tasks beat reading-only for the target words (Laufer, 2003). Weaker learners in particular struggle to pick up words from context without help.

So the failure mode is specific: a generic list, unreviewed, never retrieved, studied as one whole theme pack, and never met in real language. Flashcards built from words you actually need, reviewed on a spaced schedule, with a context line attached, do not share that failure mode. The problem is the random list, not Anki.

Practices the studies support

  • meet the word in something you actually read or hear
  • retrieve it later, from memory, on a spaced schedule
  • keep a short context or example with the word
  • do not study a whole semantic set in one sitting
  • prefer words you have a reason to need

Our article on language exchange notes uses the same sequence: catch a few words during an exchange, tidy them into fixed notes, mark the ones for next time, and reuse them. The same loop is on the home page.

Sources