Dictionaries Are Good, Actually

3b0be40258424b8fcf531a2978000780 (what is this?)

I’ve been learning Finnish for a little over a year now, and one of my favorite language-learning gamified interfaces is Clozemaster. Among many other tiny bits of brilliance, because of this:

Wiktionary lookup in Clozemaster.
Wiktionary lookup in Clozemaster.

In a tiny subwindow, Clozemaster links out to the Wiktionary page for every single word in the sentence you’re being asked for in question.

After you learn the basics of a new language, the primary challenge becomes vocabulary acquisition. Looking up a word in the dictionary isn’t nearly enough — Dr. Paul Nation estimates that you need at least 7 to 8 high quality meetings with the word in varied contexts in order for the word to really “click”:

You might hear that and think to yourself “Ah, so the dictionary isn’t really all that helpful after all”, because that’s a lot of times to see the same word either way. Not so — if you go on to 1:45, you’ll notice Dr. Nation actually thinks of learning words as a kind of conditional flow, from Noticing to Retrieval to Varied Meetings to Elaboration.

He even says outright:

We can’t do retrieval unless we’ve already met a word before and had some link with its meaning [aka, do Noticing]. (2:10)

In other words¸ it’s not surprising that you can watch hundreds of hours of anime or K-drama and walk away with a single-digit number of Japanese/Korean phrases actually grokked, because you never linked any of the words with their meanings in the first place.

As for how one should properly ‘do’ noticing, Dr. Nation recommends the following simple loop:

  1. Recognize that the word you’re meeting now is one you haven’t met before.
  2. Guess from context what it might mean.
  3. Look it up after in the dictionary to see whether your guess was correct.

In the past, step 3 took an order of magnitude more time than it does today, which is why guessing from context was such a useful skill to consciously train:

Word retrieval time elapsed (zeroth order) Parallelizable?
Recognizing a word you already know Tenths of a second Very!
Guessing what a word might mean from prefix, suffix, and root knowledge A few seconds Slightly
Looking a word up with a mouseclick A second or two No
Looking a word up with a well-loved paper dictionary Tens of seconds No
Looking a word up with a new paper dictionary A minute or two No
Getting a new paper dictionary 1 hour - 1 month, depending on historical era Yes, but irrelevant

I suspect this might be part of the reason there was such a strong emphasis on traditional grammar studies back in the days people studied Latin and Greek: The faster you get to the point where you can start to mentally break down prefixes, suffixes and word roots, the faster1 you get to the point where you can start to guess what a word means from context.

But even if you’re a fan of guessing what new words mean (and I am!) it still makes sense to look them up if they’re new to you, just to make sure. There’s just no good reason to skip the mouseclick that it takes on a site like Clozemaster, or, for websites in general, the mouseclick it takes using a plugin like Readlang. You don’t earn special points for doing things the hard way!

  1. Of course, at that point you have to start worrying about not using dream logic to link the concepts together. I imagine old medieval scholars got a lot of enjoyment out of reverse-engineering the subtle threads of conceptual webbing that were actually used to make the complicated words of Latin and Greek make sense, in a similar way to how Borges’s Pierre Menard didn’t just want to translate Don Quixote, he wanted to become the very particular kind of guy who, in 20th-century France, would spontaneously generate the exact text of Don Quixote. Kind of like how they Matrix download skills in Everything Everywhere All At Once, except instead of downloading a skill he’s downloading a really weird aesthetic. ↩︎