Grammarly has Top Linguists. Top. Linguists.

Maybe you’ve heard of Grammarly. It’s software for checking your spelling and grammar. It’s getting a lot of promotion around the internet recently, probably because it’s free and everyone is afraid of bad grammar, so I checked out their website.

The world’s leading linguists? Sounds impressive. I have complained before about grammar products created by non-linguists, so this is refreshing. Let’s have a look at the blog.

Shakespeare, punctuation tips, writing mistakes, communication faux-pas, spelling, poetry, commonly confused words….where is the grammar? OK, there’s a little bit here and there, but it doesn’t really stand out as the primary topic of the blog. It’s more about English in general. Their Twitter feed is the same thing. I like English, so that’s cool, but I just expected something run by leading linguists to have more of a focus on, I don’t know, language analysis.

To be fair, their audience probably isn’t interested in too much grammar. My guess is that most readers of the blog are not linguists themselves, but have jobs that involve writing and editing. They would prefer to read about the broader topic of communication, and have a light moment in their day, rather than go over the dark and harrowing details of sentence parsing.

After some more clicking and scrolling around, I found this:

Look at what it says under Handbook: “Detailed explanations of grammar rules”. Great. I love that stuff. This isn’t the blog, it’s the handbook, and it explicitly promises details about grammar. I expect something a little higher quality here. It should be good if this place is hiring the world’s leading linguists, right?

Sadly, no. Calling these “detailed explanations” is stretching the meanings of both those words. Many entries are no more than a paragraph in length. I picked two entries to talk about here, which are single-sentence explanations (!), padded with example sentences. The first is the entry on phrasal verbs.  For ease of reference, I’ve put up a screen caps here (luckily they don’t take up much space).

“Phrasal verbs are verbs which are used with another word (an adverb or preposition) to create a commonly used phrase.”

That’s it folks. You can go home now. There is apparently nothing more to learn about phrasal verbs.

Come on, Grammarly. Where’s the grammar? There’s nothing helpful here. There are differences between phrasal verbs, and sequences of verb + preposition. A grammar handbook should explain these differences, not gloss over them. For example, consider ‘sleep in’ in these two sentences:

  • I slept in a tent
  • I slept in this morning

The first is the verb “sleep” followed by the preposition “in”, the second is the phrasal verb “sleep in”. How can we tell? One way is by moving the in-phrase to the front of the sentence:

  • In a tent is where I slept (good; “in” is a preposition, “sleep” is a verb)
  • *In this morning is when I slept (not good; “sleep in” is a phrasal verb)

As a side note, this is technically called “clefting”, and it’s a really common sentence structure in English.  Grammarly doesn’t seem to have an entry for clefting in the handbook. Weird. If you had employed the world’s leading linguists, I’m sure that at least one of them would remind you about this topic.

Another way that you can distinguish phrasal verbs from verbs+phrases is to put words in between the verb and the preposition. You can do this if it is a verb followed by a phrase, but you can’t break up a phrasal verb:

  • I slept very poorly in a tent.
  • *I slept very poorly in this morning

I know these sentences types can look alike, but you can’t let surface similarity confuse you when analysing language. The concept of surface vs. underlying representations is literally a Linguistics 101 topic. Yet here we have a grammar handbook, purportedly written by the world’s leading linguists, that fails to grasp this elementary concept. Just who are these linguists? Did they forget to hire a syntactician somehow?

Some of the example sentences don’t even contain phrasal verbs. For instance, “know how to” in “if you don’t know how to spell”. The “to” is part of the infinitive “to spell”, it’s not a preposition, so this cannot be a phrasal verb. I’m also not sure that “check into” is a phrasal verb either, because you can cleft it (“It’s what I checked into”) and I feel like you can stick an adverb in there too (“I will be sure to check thoroughly into that”).

Grammarly gives “run away” as a phrasal verb as well. That wouldn’t count as far as I know, because ‘away’ is an adverb not a preposition. However, if you look carefully at the Grammarly definition, you’ll see they included adverbs. I’ve never seen anyone else define phrasal verbs in this way, but I guess this example is at least following their own rules.

The entry for the causative is another one-liner that manages to get everything wrong.

“Causative verbs cause an action to happen, either by allowing it to happen or by forcing it to happen. Let, make, and have are causative verbs.”

This is the entire explanation of the causative. You could at least try a little bit, Grammarly. The first sentence is also false if you take it at face value, because verbs cannot cause things to happen at all. Verbs are abstract concepts. I know that it is common short-hand to speak in these terms, so this might seem like nitpicking, but I don’t think it is acceptable in this case. It’s the only explanatory sentence! The other sentence is just a list, with no explanations. You don’t get to take a short cut on this. The icing on the cake here is that it’s a purely semantic description (which, at least, is correct). Grammarly’s grammar handbook has no grammar in this section. The world’s leading linguists work there?

For understanding the causative, it is really useful to distinguish between the subject of a verb, which is a purely grammatical relation, and the semantic role of the subject. Typically, the subject of a transitive verb has a semantic role of Actor or Agent, meaning that it does whatever action is described by the verb. When the verb is made into a causative, the subject’s role changes, and it is no longer the Actor. The subject becomes the Causer, and the role of the Actor is transferred to an new object that is introduced for this purpose. (I’ve talked about how subjects are not always Actors in some detail over here.)


  • Joe washed his car. (Not causative. Joe is the Actor – he washes.)
  • Joe had Jim wash his car. (Causative. Joe is not doing the washing, Jim is.)

Grammarly mentions three causative verbs, “let, make, have”, but their example sentences include “get” and “insist”, and no examples of “let” are given. This is not even a coherent set of verbs. Let’s start with “make, have, get”. I can agree on them being causative, but it’s important to point out that they don’t have the same grammatical behaviour. Specifically, ‘got’ requires an infinitive verb:

  • Joe made Jim wash the car.
  • Joe had Jim wash the car.
  • *Joe got Jim wash the car.
  • Joe got Jim to wash the car.
  • *Joe made Jim to wash the car.

How about the verbs “let” and “insist”? The verb ‘let’ apparently follows the same pattern as ‘make’ and ‘have’:

  • Joe let Jim wash the car.

However, this sentence does not have a causative interpretation. It means Jim asked, and Joe granted permission. No one was forced or obliged to do anything here, so it’s a real stretch to call it causative. It’s also possible that Grammarly meant this kind of “let”:

  • Let him wash the car!
  • Let’s wash the car!

This would also not be a causative. This is technically called a “hortative“, which is a subtype of imperative.

The verb “insist” also seems to follow the causative pattern, but actually it is just a regular verb that takes a subordinate clause in the subjunctive mood. The key piece of evidence here is that it’s possible to introduce the relative clause marker ‘that’ after the verb “insist”, but you can’t do that after a true causative:

  • Joe insisted Jim wash the car.
  • Joe insisted that Jim wash the car.
  • Joe made Jim wash the car.
  • *Joe made that Jim wash the car.

The second way you can tell the difference is through the semantics. With “insist”, as with “let”, the activity in question isn’t necessarily completed. The causative, on the other hand, entails that the activity was completed, and sentences sound strange if you try to cancel the entailment:

  • Joe insisted that Jim wash the car, but Jim never did it.
  • Joe let Jim wash the car, but Jim never did it.
  • *? Joe had Jim wash the car, but Jim never did it.
  • *? Joe made Jim wash the car, but Jim never did it.

All of this makes me wonder: does Grammarly’s software work? The author of the handbook obviously has no background in linguistics, and has a very limited ability to formally describe English. This doesn’t give me much confidence that the team behind this project could teach a computer anything about grammar. On the other hand, there are a lot of positive reviews, so apparently it does the job. Maybe my question should really be: why does Grammarly’s software work? The spell-checking aspect of Grammarly does not require any sophisticated knowledge of linguistics, of course. The grammar-checker does though.

One possibility is that Grammarly is just a giant list of Things To Avoid In Writing and an equally big list of Possible Solutions to Problems.  It looks for pre-defined patterns that are considered “bad grammar”, and then suggests possible fixes. It doesn’t have any real ability to analyse a text. Certainly it would be impossible to teach a computer to identify a causative construction based solely on the handbook entry.

Another possibility is that the software and the website are run by completely independent teams. The competent linguists are working on the software, and the website content is written by people who don’t have a background in formal language analysis.

After some more poking around their website, it seems that the real situation is somewhere in between. Grammarly has two offices, one in San Francisco, USA and one in Kiev, Ukraine. The Ukrainian team is responsible for the software aspect and according to the Jobs page, they are looking for people with advanced degrees in computer science. One of the job ads starts with “Do you read research papers about kernel methods for fun? Are you dreaming about gradient descent when you sleep?”. In other words, they aren’t hiring people with a passion for language or linguistics, they’re hiring mathematicians and programmers (aka “computational linguists”).

The website content creation is left to people with other backgrounds. The job for Managing Editor, for example, requires a B.A. in a “writing-intensive” program, with English or Journalism preferred. Of course, writers and programmers are vital to this enterprise, but it’s still weird that they never considered hiring someone who has training in grammatical analysis. And that’s why their Handbook sucks.

If there really are some of the world’s leading linguists working over at Grammarly, then please, please, please, get that handbook edited. And if you don’t have anyone on staff right now who can update the handbook for you, well, I have a Contact page :-)

Update June 3, 2015: Grammarly has contacted me on Twitter to say that they are currently in the process of fixing the handbook. Let’s just wait and see!



Filed under Book Review, Linguistics

3 responses to “Grammarly has Top Linguists. Top. Linguists.

  1. Heh! “The web” is often short for “the web of lies.” Betcha there was no list of those top linguists. (Maybe they hired people who were good at making linguini.)

    Liked by 1 person

  2. Well that’s frustrating. Come on, Grammarly, you can do better!

    Liked by 1 person

  3. Pingback: Grammarly doesn’t understand the subjunctive | linʛuischtick

Tell the world what you think!

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s