Adverbs? Evil, or Necessary Evil?

My WIP is still with the editor, who is very thoroughly taking notes about how poorly I represented disadvantaged groups, and otherwise pillaged the English language for the shiniest words before torching the rest.

Meanwhile, one of my beta readers (you know who you are), has repeatedly suggested to me that I kill all my adverbs. This is apparently a school of thought. Not being a professional writer, or a trained writer, this is something I had never encountered before.

Whelp, I wrote nearly 90,000 words all chock-a-block with adverbs, so you can imagine my discontent at this notion.

Then, this weekend, I ran into this post on Facebook. Basically someone else questions this mantra.

I respond to this like I do every challenge. WITH DATA! I find the complete Brown Corpus, do a sloppy adverb search (for every world ending in “ly”), and crunch percentages.

So the Brown corpus. It was assembled million words (1,014,940 according to regex) of text designed to be statistically representative of English language. It has 14,688 adverbs in it (words ending in ly excluding “only”).

This is 1.45% of the corpus.

Oh boy, moment of truth.

My WIP has 1546 adverbs (same criteria), out of 84,959 words, for a [drumroll, please] of 1.82%.

Is that significant? Let’s do some stats.

Yikes, it is. The margin of error around my percent is 0.09%. By Z-testing, the difference is significant to a p value of 0.0004.

Let’s flip this around. By putting in “the correct” proportion of 1.44%, my WIP should have no more than 1230 adverbs. Let’s add in the margin of error, for 1250. That means I have 300 more adverbs than I “should”?

Is that too many adverbs?

But then, I found a version of the Brown Corpus that had parts of speech tagged. That meant my sloppy methodology could be tightened up. So I indexed every word, along with its part of speech.

But now I had another problem. I couldn’t just search my WIP for “ly”. I had to have it tagged. Luckily, after some flailing, I found the Stanford tagger, which did just that.

So with my text tagged, I tore it apart, and analyzed it.

The Brown Corpus: 37,863 adverbs out of 1,002,550, or 3.8%.

My WIP: 6,074 adverbs out of 87,945 words, or 6.9%.

Oh, no! The margin of error was 0.2%. Yeah, I’ve got twice as many adverbs as the Brown corpus.

Okay, time to settle down. Anybody with a laptop can plunk some numbers into Excel. It takes a REAL DATA SCIENTIST to lie with it.

So the Brown Corpus is composed of many different types of text samples. News articles, scientific journals, and government publications. And also, of course, fiction. Maybe if I confine comparison to only fiction?

So 24% of the corpus, by words, was either Adventure, Fiction, Mystery, Romance, or Science Fiction. That is 234,500 words. Of which 11,417 are adverbs. Or 4.9%,

Sigh. Maybe I’m not the data scientist I thought I was. Looks like I need to cull 1,792 adverbs from my manuscript.