I just found the 'word frequency' tool in Scrivener's text statistics, but so far all I've done with it is read down the list (it occasionally generates amusing sentences) and check that words like "puissant" haven't started multiplying.
Is there a way to use word frequency data to improve my writing?
How to evaluate repeat words:
1. Not all frequencies are created equal.
a, an, the, he said, she said etc ... <- These are fine. They are considered invisible. You can prune them out if you like, but do not need to necessarily.
2. However. Some common words that tend towards becoming frequent, are more problematic.
heard, look, thought/think, smile, nod, knew, understood, etc ... <- these types of words are used heavily by beginners, because they're quick and easy. But, you end up with bobbing heads, smiling at each other, and it's neither realistic nor enjoyable to read about for very long.
2a. Some of these words are simply shortcuts. 'Smile' and 'nod' are examples of shortcuts. In my opinion, these are the written equivalent of an author stammering. (I have too many 'smiles' in my story at the moment.) To say a character 'understood something' or 'realized something' may indicate laziness on the author's part (a shortcut). (It's also telling instead of showing.)
2b. A subset of these are something called filters. Filters are those physical actions that a character does to experience their world.
I can write following, which has two filters (bolded):
Janice heard a car pulling into the driveway. She knew they were here.
Assuming we are in Janice's point of view, we can skip the "Janice heard" and 'She knew.' It goes without saying. We can play with the actual sound instead:
Gravel crunched under tires, a slow sickening sound. They were here.
So, the words heard and knew are filters, because they put a layer between the story and the reader, a bad thing. Sometimes it is appropriate to use these words, but not always. Look for the word frequency of shortcut words and filters.
3. The same thing happens with phrases and gestures. For example: Looking out of windows, pouring some sort of drink, falling to sleep <- These are phrases that are essentially just filling space, giving the characters something meaninglesss to do.
They might be better than nothing, but too much of any of these sorts of meaningless actions becomes a problem.
4. 'ly adverbs' (and too many adjectives) - If scrivener allows, see if you can find all adverbs or all 'ly adverbs.' Also look for adjectives. Yes, we want some, but if you've ever read something with too many you know how annoying they become. Readers lose the story. This is less a case of repeating words, and more a case of relying too heavily on a particular part of speech. (Adverbs can also be seen as a sort of shortcut.)
So, how to use the frequency tool? Scan through your results. If you see something in the list that you didn't realize you were using so heavily, check your document. Play with alternatives for those words. Some can be deleted straight out.
The worst repeats are the ones in the same paragraphs, because the reader brain remembers what was earlier in the paragraph. And so if you use a word later in the paragraph that was also used earlier in the paragraph, it will start to pull the reader out of the story. By the end of the paragraph they may want to throw the book across the room.
You have an excellent answer from @DPT for what to do with the results. Just a few tips that are Scrivener specific:
In the Word Frequency window, you can click on Count to sort them into numerical order. This will help to put all those words like 'I', 'it', 'and', 'she', and so on, to the bottom and allow you to start with the less frequently used words.
Still scan these single words. You can spot spelling mistakes here that you may have missed during the edit.
In the latest version of Scrivener (not sure about earlier versions) you can use the Options in Statistics, and under Word Frequency Options, 'Set List of Words to Ignore' and in there you can add the words mentioned above so they never come up in the stats to begin with.
Finally, I agree with @J.G., use the stats to find and replace obscure words. You're trying to create a picture in the reader's mind. Obscure words muddy the image. For a great example of a literary masterpiece that doesn't resort to anything even close to purple prose to create mastery, read The End of The Affair.
