How many different words are in the average novel?

Question

I realize there's going to be a bit of variation here from one author to the next and depending on length of the novel, but the range shouldn't really be too large.

The thing is, I really have no idea where the middle of that range lies. 5,000 words? 10,000? 20,000? I know the average adult vocabulary (in English) ranges between 20,000 and 30,000 words, although I don't know if that number includes variations (like, likes, liked, liking), but I would expect this number to be significantly less than the full breadth of a person's vocabulary.

I had trouble finding any information. Is there anyone who is familiar with this subject?

Answer 1

You know that you have an important question when your topic already has a recognised acronym NDW.

However it is typically used in the medical field rather creative writing. It appears to have arisen from the rehabilitation of patients with brain injury such as stroke.

The University of Albert Department of Linguistics has a fairly extensive discussion on using NDW as one measure (among many) in interpreting the complexity of texts.

And the short answer to your good question: IT DEPENDS on a huge range of factors. Any "average" value is likely to be meaningful only for the sample from which it was calculated. You could calculate an NDW for Stephen King novels but that would have little value in making predictions about the work of Umberto Eco.

A concordance of the Bible gives an interesting way to examine some of the issues. (Although this is KJV. I would be interersted to see the differences in GNT.)

Answer 2

This is almost impossible to answer with any degree of accuracy as it depends on so many variables. However, as a very rough and general guide the average novel length is about 80-100,000 words of which about 5-10,000 words would be unique and differentiated.

Answer 3

While I can't answer the "big" question of an average, I can answer the smaller question of whether anyone's heard/seen anything on the subject before. If anyone has answered a question like this I would assume it would be Franco Moretti or someone doing similar computational research into the statistical profiles of large literary corpora.

Wired did an article on Moretti's research a few years ago. He floats a concept of "distant reading" wherein we cover a lot of ground traditional criticism can't meaningfully cover through statistical methods. I haven't read him widely, so it's possible he's covered this somewhere. I've emailed him before and he was responsive to my questions, though I note from his faculty page that he's on leave this year which may have some effect (as might any pending publications he has on the topic). He also has related books, like Graphs, Maps, Trees & Distant Reading.