Grammar checkers vs human judgment: Still no contest

Every writer I know (at least, the ones with whom this topic has come up) turns the grammar check off on his or her computer. We all believe that our own feel for grammar is way better than that of whoever programmed the grammar check for Word. And we’re all right, too. (Spellcheck is a different story; that’s a really handy little gizmo.)

But recently I’ve been playing around with a new grammar check service,, because I got an offer to play with it for free. Plus, I was thinking about maybe pointing this service out to some of the students who come in for help with their papers, and maybe to the English faculty – the students because the better ones could often use a boost with grammar but shouldn’t be overwhelmed by this service, and the faculty because not only generates summaries of student errors, but also checks for signs of plagiarism.

I wasn’t really thinking of this service for the use of writers, because, well, if you’re a writer, then the English language is your one and only tool, right? And if you have no feel for it, good luck with that. So you are not likely to need a grammar checker if you’re a writer. Unless! If grammarly could catch certain really obnoxious mistakes, and not annoy you by pointing out too many perfectly fine constructions by mistake, then maybe it would be worthwhile? Depending on your style, of course. Because probably you have noticed that novels, like blog posts, may be deliberately written in a more casual style with deliberate errors, like for example in this paragraph?

So I tried three documents: a horrible, mistake-laden student-type paper I got off the internet (I did clean it up just a little before I ran it through grammarly, because it was actually too awful to use straight), a handout I wrote years ago on how to housetrain a puppy, and a six-page excerpt from BLACK DOG, which as you all know is the novel I have coming out next spring.

So What Did I Find?

Wow, did grammarly nail that downloaded student paper for plagiarism. Since in fact I did get the whole thing straight off the internet, it certainly would have been plagiarism if I’d meant, God help me, to turn it in for a class. Grammarly also found zillions of mistakes in the paper, but I didn’t want to analyze what it was doing because that paper turned out to be just too painful to read closely, even in the interests of analyzing grammar checkers.

As for the handout and the novel fragment, I found that grammarly can be set for “casual” and “creative” as well as “academic”, and if you use those settings, it doesn’t question as many sentences. This is great! But it doesn’t go far enough. It identified as wrong four types of constructions that I was using on purpose, in each case because it wanted me to write more formally.

Places Where Grammarly Should Really Lighten Up

a) Grammarly thinks it’s wrong to split infinitives. We all know split infinitives are generally accepted today, except in extremely formal academic writing, and that you not only can but should split ’em if you think the sentence sounds better that way.

b) Similarly, grammarly thinks prepositions are items you shouldn’t end a sentence with, and in fact often this does feel clumsy, imo. But, especially in slang or in dialogue, sometimes it sounds best to end a clause or sentence with a preposition: “Hey, I’ve got a hot date tonight!” “Dude! Who with?” is not an exchange to re-write according to formal rules that are outdated anyway. And I say this as a person who has a “I’m the grammarian about whom your mother warned you” tee-shirt.

c) Grammarly doesn’t like constructions such as “waving away any concern” and “put back her hood.” Given the little grammar tips grammarly showed me, it wanted me to write “waving any concern away” and “put her hood back.” But those corrected phrases sound clumsy in context, so I think that whatever supposed rule governs this is at best questionable, or at least being interpreted too strictly.

d) Grammarly hates partial comparisons, such as “People eat more in the cold.” I gather it wants complete comparisons, such as “People eat more in the cold than they do in warm environments.” I get what it’s talking about here, but can you imagine writing dialogue like that? Honestly, you’d get an instant Mr. Spock impression.

Other Stuff I Noticed About Grammarly

I’ll start by saying that grammarly identified all the Spanish words in BLACK DOG as misspelled, which you can hardly blame it for, so I didn’t count that as a problem. Besides spelling and the four issues mentioned above, grammarly identified 24 “questionable writing issues.”

a) It identified seven commas as misused. Four of those sentences were exactly the way I want them, one definitely correct – granted, the sentence structure was complicated. The other three are commas that a copy editor also might mark, and that I’d think hard about. It’s a question of how the sentence sounds in dialogue (or direct thoughts, which are like dialogue), and whether I want a pause or whether (in the case of a missing comma) I want a sense of rushed speed through that sentence. Most important would be thinking, in every single case, about whether I think the way the commas are used might conceivably distract a reader. Commas are tricky. I think grammarly was right to mark the ones it did, even if I chose to leave every single one alone.

b) Commonly confused words. It marked five, all of which were correct the way I had them. In every case but one, the suggested correction was actually ridiculous. Honestly, I am not very likely to type “bony” when I mean “body”, or “pups” if I meant “pumps.” Unless I’m really tired. (Then I quit trying to write and go to bed.) IfI used grammarly on a regular basis, I would want a way to turn this particular function off.

c) Grammarly marked five possible verb-tense errors. It was just plain wrong in every case – wrong and funny in one case when it thought “raven” was a verb and suggested “ravening” or “ravened.” I’d meant the bird, of course!

d) And last, grammarly marked a handful of possible errors with articles, all of which were fine the way I had them. When someone says, “Be patient,” you don’t need an article in front of “patient!” Understandable for grammarly to flag this, and yet also stupid. Also, grammarly wanted me to put an article in front of the word “something” – as in “a something” or “the something.” This was almost as funny as the raven.

Things I Wish Grammar Checks Would Catch, But They Just Don’t:

Grammarly did no better than Word’s standard grammar check in catching repeated words. I put the word “steep” in the second paragraph three times, just to see if grammarly might by any chance be programmed to flag this kind of thing. It didn’t. Too bad! Copy editors do a great job with this, but it’s amazing how something like this will sneak right through into the final version of a manuscript, despite everything you can do.

Grammarly also did not catch an incorrect “him and I” construction, which I put in because I really hoped it might. I know the ordinary Word grammar check doesn’t, but I hoped grammarly would do better. I don’t know what it is with the people who program grammar checks, I really don’t. Were ALL of them so traumatized by their grade school teachers that they are scared to put the word “me” in a sentence? (As a related question, why is the one rule that sticks FOREVER in peoples’ heads a rule which doesn’t even exist?)

Grammar checkers do get the difference between subjects and direct objects; I’m sure they’d mark a sentence like, “Steve was so mean to I,” so why would a grammar checker let you get away with “Steve was so mean to Melissa and I”? For all I know, a grammar check might even INTRODUCE this kind of error, since plainly grammar checkers think the “him and I” construction is just peachy.

Okay, winding up: it took grammarly 40 seconds to analyze six pages. (Yes, I timed it.) I have no idea how long it would take to check a whole novel, but . . . quite a while, I bet.

Also, while grammarly generated a printable report which flagged everything and suggested corrections, the report was generated in a single-spaced all-one-paragraph format which would be unimaginably painful to read if you had more than a few pages. Though I guess it would save paper, if that was a priority.

Who Would Benefit From a Service Like Grammarly?

Students, maybe. Not in basic writing, but maybe at about the English Comp II level or above. If students are truly clueless about grammar, this program wouldn’t work well. You have to have some basic sense of what you’re doing before you can evaluate whether you did indeed put a comma at the right place, when a program suggests you didn’t. For students at too elementary a level, I think this program would tempt them to change constructions that are right into constructions that are actually wrong. But for students who generally write well, but who tend to make a handful of standard errors, I bet this program would be helpful in teaching them not to. Unless the errors were “him and I” mistakes, in which case, sorry, but it would miss ’em all.

Teachers, maybe. You sure wouldn’t want to have to wade through every single thing grammarly flags as a possible mistake, but it might be quite handy to have grammarly’s quick summary of potential areas of confusion. Grammarly does generate a summary of what it thinks might be spelling errors, commonly confused words, article errors, pronoun errors, adjective and adverb errors, faulty parallelism, confusing modifiers, verb-form errors, comma errors, etc. Even though grammarly would probably be frequently wrong about verb tense and article usage and whatever, that kind of summary still might be useful to you a better sense about where a particular student is having trouble. Plus, flagging possible plagiarism is always good.

Writers? In my opinion, no. Anybody who seriously wants to write ought to be too fluent in the language to benefit from this or any grammar checker – except for checkers with actual judgment, such as beta readers and copy editors. Those grammar checkers are priceless.

Please Feel Free to Share:


Leave a Comment

Your email address will not be published. Required fields are marked *

Scroll to Top