How many words you need to learn to speak a foreign language? Many, but you can already do a lot with a fairly limited vocabulary. The problem is that the words you don't know will often be quite important to understanding the whole text or conversation.

I have taken two short texts and grayed out rarer words - that is, those not in the 3000 and 5000 most common words in English. That will show you the difference between reading with a 3000-words-vocabulary, with a 5000-words-vocabulary and with a perfect vocabulary.

Newswire article'Chinese is the Language to Learn' (basic journalistic prose)
• With a 3000-words-vocabulary
• With a 5000-words-vocabulary
• With an unlimited vocabulary

Book extract'How to Learn any Language' (basic litterary prose)
• With a 3000-words-vocabulary
• With a 5000-words-vocabulary
• With an unlimited vocabulary

Methodology
Obviously this makes the implicit assumption that you will learn words according to their frequency, that all foreign language material you will be confronted to has the same frequency distribution and we omit the fact that you may understand 'unknown' words through roots or cognates. A point commonly made in such discussions is that even though you may understand 80# of any texts with only 3000 words or so, to cover the remaining 20# you do need many more words. More interestingly is to wonder how much you miss of the text's overall meaning by losing those 20#. See for yourself in the example above.

For a long time I wanted to do this article but could never find a lemmatized frequency dictionary of English with more than 6000 entries. Finally I decided to make do with Adam Kilgarriff's and its 6318 entries (if a word can be a noun and a verb, it is counted as two entries I'm afraid). Anybody who can get me a 20,000+ lemmatized frequency dictionary for English gets a free chocolate box.

I took a fairly simple 'litterary' text from 'How to Learn Any Language' by Barry Farber, the best book I know about language learning. Few people would call it 'litterature' and yet it uses enough rarer word for me to make my point.