How many words do you need to have read?

Lately I've been thinking a lot about the need for extensive reading, and how that's challenging for Ancient Greek, but also very necessary. In particular, the question I've been asking, or set of questions, looks like the following:

1. How many 'words' do you need to 'know' to reach certain proficiency milestones in Ancient Greek?
2. How many exposures to a word do you need to (a) passively recognise a word, (b) actively recall a word?
3. How many words do you need to read to reach those exposure-points for certain milestones?

If you're familiar with this kind of work, you'll know that it rests solidly on the extensive research of Paul Nation in this area. That's where I'm starting from, and I'm thinking about how this applies both broadly and narrowly, in terms of grading reading programs, and total words vs words known.

Whenever I have a question like this, I always figure that the best person to ask is James Tauber, because if he doesn't know the answer, either he knows who does, or how to calculate it, or most often "I was working on this X years ago, here's an old post about it."

Which is exactly the case here, in fact he and I discussed this 7 years' ago. The numbers differ across different corpora, but if I summarise some of the findings from that earlier work:

For the New Testament, you cross 95% at 2000 lemmas, 98% around 3000.

For all of Perseus / OGL in 2019, it's 95% at somewhere between 4-5000 lemmas, 98% at 8000 lemmas.

There's a very, very long tail in Greek literature, which means that you never stop meeting new words, The fact that it's 8k lemmas at 98.00% but then 20k at 99.75% tells you how you'll never quite get there! (

Nation does has a an attempt to correlate CEFR levels and Vocabulary sizes, which you can access on this page. There he also cites some research. Just looking at Nation's suggested vocabulary sizes, and noting that he is estimating on the high side compared to others, I think that is probably reasonably comparable to Ancient Greek. Whatever morphological savings you get from a richly inflected language, seem to be somewhat cancelled out by the breadth of Ancient Greek words.

To replicate Nation's table in brief:

A1 - 120 survival words

A2 - Most frequent 1000 word families

B1 - 2-3000k

B2 - 4k

C1 - 5-6k

C2 - 7-9k

But what do you need to learn that many words? Nation's 2014 article, 'How much input do you need to learn the most frequent 9,000 words?' attempts to answer this.

Using 'buckets' of 1000 words, and corpus data, with the assumption that you need at least 12 repetitions, Nation calculates:

2k words: 171,411.

3k words: 300,219

4k words: 534, 697

5k words: 1,061,382.

6k words: 21,450,068.

7k words: 2,035,809.

8k words: 2,427,807.

9k words: 2,956,908.

So, if you want to get to the point of reading almost all of Greek literature with 98% vocabulary comprehension, you need to also probably read around 3 million words of Ancient Greek. But even to get to 4-5k lemmas, where you would have 95% coverage, you need 500k-1 million words of reading.

Here is the question I keep asking myself as a result of all this: how to provide scaffolded and graded reading to help students read the first 100k, 200k, 300k to build their vocabulary from zero to 'I can read authentic ancient texts without much help'?

More on this in later posts.