An English PhD-Holder Reviews Grammarly

I’ve been resistant to software text editing tools for a long time.

Partly that’s because my PhD is in English, and editing text is something I already do reasonably well. The idea of a software package accomplishing one of the few tasks I am qualified to perform is a touch unsettling.

But the main reason I don’t like text editing software is that it’s always seemed like a bit of a blunt instrument. Sure, these tools might pick up obvious misspellings and grammatical errors. But no software package is going to understand the subtler aspects of language: tone, genre, dialect, sarcasm, wordplay, and so on.

The wiggly red and blue lines in Microsoft Word are a perfect example of an infuriating text editing tool. I get those wiggly lines on writing that is actually correct for my purposes; it’s just non-standard as far as a bunch of dumb programming can determine.

For instance, unless I’ve ‘added to dictionary’ in that particular licensed copy, MS Word always tells me that my own name is a spelling mistake. (Horrifying-but-true sidenote: it also suggests that I correct ‘Anaise’ to ‘Anus’. This has left me with an unreasonable hatred of Bill Gates.)

It’s not just MS Word that has limitations, either. Even writing tools with good academic credentials, like Helen Sword’s The Writer’s Diet test, are limited to sentence-level analysis. Software, according to every teacher I’ve ever had in an English department, simply isn’t as sophisticated as a human reader.

So it has taken me ages to try Grammarly, and I did so with a lot of cynicism.

But part of my job is to support people who write theses and dissertations, and many thesis-writers consider using Grammarly. So out of professional curiosity, I downloaded the free MS Word add-in and had a go.

Testing on a Short Piece of Writing

First, I tested Grammarly by using it to check a Thesislink post that I wrote last week. This was a published post, and I had proofed it myself (the old-school way) before publishing it, so I thought the writing was of a good standard. I copied the text over to MS Word, fired up the Grammarly add-in, and had it scan my work.

Grammarly found 6 critical issues. Of those, I agreed immediately with one. I’d written that something was “testament to” when I should have written “a testament to.” Good catch, Grammarly!

But the other 5 ‘critical errors’ weren’t so straight-forward. One of Grammarly’s identified ‘errors’ was actually correct in the context of the sentence. Others were technically examples of imperfect English, but I would deem them acceptable for the conversational style of a blog post. Bottom line: if I had blindly accepted all of Grammarly’s suggested changes, my writing would have been marginally worse for its intended purpose, not better.

Of course, the great thing about text editing software is that you don’t have to (in fact, you shouldn’t) blindly accept all suggested changes. If you did, your writing would probably end up looking robotic. By selectively accepting some of the suggested changes, you can hold on to your stylistic choices while still getting rid of genuine errors.

After accepting one of the suggested changes and considering others, I was able to make the piece better than it had been after a simple proofread.

Testing on a Whole Thesis

So Grammarly did well enough in a simple test. But to really put it through its paces, I used Grammarly to check my whole PhD thesis. All 90,000 words of it.

This was a serious test. My thesis was proofed not only by me (many, many times) but also by one of New Zealand’s leading English professors, who mercilessly circled hundreds of errors — however slight — in rings of red pen. It was passed by a panel of international examiners with no suggested revisions and was even nominated for an award. I don’t tell you this to brag; I tell you this to illustrate that, after much careful proofing and editing, the standard of writing in this thesis was very high.

Grammarly reviewed the whole thing in about 30 seconds and reported a horrifying 713 critical issues. Over seven hundred? That couldn’t be right.

It got worse from there. When using the free version, Grammarly gives you this annoying message:

Oh, the anxiety! You’re telling me that there are 1519 other writing issues in my thesis and I have to pay to find out what they are?! This will haunt me.

That little box is a very clever marketing ploy. The premium version of Grammarly claims to identify opportunities for ‘vocabulary enhancement’ and identification of ‘advanced issues’ – but exactly what those issues might be isn’t clear until you pay a minimum of US$11.66 (approx NZ$17.92) per month. If you plan to buy short-term access for reviewing your final draft, the price is US$29.95 (approx NZ$46.02) for a one-month pass.

I’m a cheapskate, so I didn’t buy the paid version. That left me with only the free version’s 713 critical issues to review. Still: seven hundred and freaking thirteen issues? Thankfully, Grammarly wasn’t correct about the scale of the problems in my writing. But it wasn’t all wrong, either.

So what was behind all these ‘critical issues’? Well, for a start, Grammarly is not so good at making allowances for words with multiple meanings. In one sentence in my thesis, I wrote about growing a plant “from a cutting.” Grammarly interpreted the word ‘cutting’ as a verb, and suggested omitting the ‘a.’ In fact, I was using the word ‘cutting’ as a noun (meaning a piece cut from a plant). Grammarly’s suggestion there was simply wrong.

Quoted material was another source of many of Grammarly’s false errors. I quoted extensively in my thesis, and I quoted carefully. Where I quoted international authors, I kept their international spelling; where I quoted obscure academic terms or names, I transcribed them precisely. Boom. Hundreds of ‘errors’ that were actually, in context, just correct quotes.

Grammarly’s take on subject-verb agreement is also a little unsophisticated. In my introduction, I wrote that “the manipulation and replication of genetic material has occurred for millennia.” Grammarly identified ‘has’ as an error, and I can understand why. ‘Has’ is used for singular subjects (my cat has fur) whereas ‘have’ is used for plural subjects (my cats have fur). Because I appeared to have listed two subjects (“manipulation and replication”), Grammarly wanted me to use the plural verb ‘have.’ But in the context of the sentence, my subject was actually singular since both “manipulation” and “replication” were part of a single subject phrase describing genetic alterations. Therefore, ‘has’ was arguably the correct choice.

Grammarly also threw up a lot of false errors in my bibliography. For instance, a lot of authors’ names were identified as ‘spelling mistakes.’ Also, Grammarly doesn’t cope well with the types of capitalization required in referencing styles. And perhaps unsurprisingly, ‘doi’ comes up as an unknown word. (It’s easy enough to add those unknown words to your Grammarly dictionary, just as you would with MS Word; so this isn’t a deal-breaker.)

But Grammarly did get a few things right. It correctly identified that I should not have added periods after every instance of the title ‘Dr’. That would be correct in American English, but not in my New Zealand English. Grammarly also correctly identified my habitual sloppiness with commas.

Grammarly even found a few non-obvious spelling errors that had been missed by me, my English professor supervisor, and two examiners.

Overall, I’d estimate that 70% of the errors Grammarly identified in my thesis were false. Another big chunk of the errors were writing features that were technically incorrect but made sense in context. A small minority — I’d guess about 3% — were things I wish I could change.

With 713 critical issues, even if only 3% were legitimate errors, that’s still a good 20 errors in my polished, proofread thesis! Twenty errors that I made, and didn’t correct; twenty errors that my red pen-wielding supervisor didn’t spot. And that’s just what the free version of Grammarly picked up – I hate to think what the premium version found among my 1519 “additional writing issues.”

Is Grammarly Worth Using for Your Thesis?

Yes! My cynical heart has been warmed. The free version of Grammarly does a surprisingly sophisticated job of picking up a broad range of errors. However: user beware. It also picks up a lot of false errors. It is only a software tool after all, and it doesn’t read human writing the way a human does. Do not — I repeat, DO NOT — take all of its suggestions.

One thing to consider before using Grammarly on your thesis, though, is your privacy. Grammarly’s Privacy Policy allows their employees to review your writing for a number of reasons, including to improve their algorithms. If your writing contains sensitive information, that may be off-putting.

You should also know that Grammarly doesn’t allow you to set your language to New Zealand English. British and Australian English are the closest options, and for most NZ thesis-writers they will probably suffice.

With all that being said, I think Grammarly is a fantastic tool to use for a ‘first sweep.’ Run your writing through it before you send it to a supervisor. Clean up the obvious errors so that you (and your supervisor) don’t waste time on basic corrections. The free version of Grammarly is fine for that.

But the real work of editing is in understanding the subtleties of an individual piece of writing, and working to preserve and clarify its meaning within accepted standards for its genre. Leave that work to humans: yourself, your supervisors, and any professional proofreaders you may like to hire. Because that work is too important for a software tool.

I hope so, at least. Otherwise, my degree is kind of redundant.

Note: The folks at Grammarly didn’t ask me or pay me to write this review. In fact, I’ve never had any contact with them at all.

About Anaise Irvine

Dr Anaise Irvine is the Editor of Thesislink and leads the Researcher Education and Development team at Auckland University of Technology. Her PhD research analysed how contemporary films and novels represent genetic engineering as a social justice issue. These days she works with researchers at all levels to improve their research skills, and the most obscure of her own research skills is being able to turn novels into phylogenetic trees!

Leave a Reply

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

 characters available