"Hemingway" on Hemingway

« previous post | next post »

Several people have written to me about the so-called "Hemingway" app, which offers to give you detailed stylistic advice about your writing.  One useful way to evaluate programs of this kind is to see what they do with good writing — and given this effort's name, it makes sense to check out its opinion about the prose of Ernest Hemingway.

Here's the first paragraph of Hemingway's 1923 story My Old Man:

I guess looking at it now my old man was cut out for a fat guy, one of those regular little roly fat guys you see around, but he sure never got that way, except a little toward the last, and then it wasn't his fault, he was riding over the jumps only and he could afford to carry plenty of weight then. I remember the way he'd pull on a rubber shirt over a couple of jerseys and a big sweat shirt over that, and get me to run with him in the forenoon in the hot sun. He'd have, maybe, taken a trial trip with one of Razzo's skins early in the morning after just getting in from Torino at four o'clock in the morning and beating it out to the stables in a cab and then with the dew all over everything and the sun just starting to get going, I'd help him pull off his boots and he'd get into a pair of sneakers and all these sweaters and we'd start out.

Pseudo-Hemingway's evaluation: "Bad". The alleged problems:

1 of 3 sentences are hard to read.
2 of 3 sentences are very hard to read.
1 adverbs. Aim for 0 or fewer.

Here's the start of The Snows of Kilimanjaro (minus some dialogue):

The cot the man lay on was in the wide shade of a mimosa tree and as he looked out past the shade onto the glare of the plain there were three of the big birds squatted obscenely, while in the sky a dozen more sailed, making quick-moving shadows as they passed.

He lay then and was quiet for a while and looked across the heat shimmer of the plain to the edge of the bush. There were a few Tommies that showed minute and white against the yellow and, far off, he saw a herd of zebra, white against the green of the bush. This was a pleasant camp under big trees against a hill, with good water, and close by, a nearly dry water hole where sand grouse flighted in the mornings.

Pseudo-Hemingway's evaluation: "OK". The problems:

 2 of 4 sentences are hard to read.
1 of 4 sentences are very hard to read.
2 adverbs. Aim for 1 or fewer.

A paragraph from a bit later in the same work:

That was the day he'd first seen dead men wearing white ballet skirts and upturned shoes with pompons on them. The Turks had come steadily and lumpily and he had seen the skirted men running and the of ficers shooting into them and running then themselves and he and the British observer had run too until his lungs ached and his mouth was full of the taste of pennies and they stopped behind some rocks and there were the Turks coming as lumpily as ever. Later he had seen the things that he could never think of and later still he had seen much worse. So when he got back to Paris that time he could not talk about it or stand to have it mentioned. And there in the cafe as he passed was that American poet with a pile of saucers in front of him and a stupid look on his potato face talking about the Dada movement with a Roumanian who said his name was Tristan Tzara, who always wore a monocle and had a headache, and, back at the apartment with his wife that now he loved again, the quarrel all over, the madness all over, glad to be home, the office sent his mail up to the flat. So then the letter in answer to the one he'd written came in on a platter one morning and when he saw the hand writing he went cold all over and tried to slip the letter underneath another. But his wife said, "Who is that letter from, dear?" and that was the end of the beginning of that.

Pseudo-Hemingway's evaluation: "Bad". The problems:

3 of 7 sentences are very hard to read.
3 adverbs. Aim for 0 or fewer.

The start of For Whom the Bell Tolls (again minus the dialogue):

He lay flat on the brown, pine-needled floor of the forest, his chin on his folded arms, and high overhead the wind blew in the tops of the pine trees. The mountainside sloped gently where he lay; but below it was steep and he could see the dark of the oiled road winding through the pass. There was a stream alongside the road and far down the pass he saw a mill beside the stream and the falling water of the dam, white in the summer sunlight.

He spread the photostated military map out on the forest floor and looked at it carefully. The old man looked over his shoulder. He was a short and solid old man in a black peasant’s smock and gray iron-stiff trousers and he wore rope-soled shoes. He was breathing heavily from the climb and his hand rested on one of the two heavy packs they had been carrying.

Bending under the weight of the packs, sweating, they climbed steadily in the pine forest that covered the mountainside. There was no trail that the young man could see, but they were working up and around the face of the mountain and now they crossed a small stream and the old man went steadily on ahead up the edge of the rocky stream bed. The climbing now was steeper and more difficult, until finally the stream seemed to drop down over the edge of a smooth granite ledge that rose above them and the old man waited at the foot of the ledge for the young man to come up to him.

Pseudo-Hemingway's evaluation: "OK". The problems:

3 of 10 sentences are hard to read.
2 of 10 sentences are very hard to read.
5 adverbs. Aim for 1 or fewer.

The first paragraph of The Old Man and the Sea:

He was an old man who fished alone in a skiff in the Gulf Stream and he had gone eighty-four days now without taking a fish. In the first forty days a boy had been with him. But after forty days without a fish the boy's parents had told him that the old man was now definitely and finally salao, which is the worst form of unlucky, and the boy had gone at their orders in another boat which caught three good fish the first week. It made the boy sad to see the old man come in each day with his skiff empty and he always went down to help him carry either the coiled lines or the gaff and harpoon and the sail that was furled around the mast. The sail was patched with flour sacks and, furled, it looked the the flag of permanent defeat.

Pseudo-Hemingway's evaluation: "OK". The problems:

2 of five sentences are very hard to read.
2 uses of passive voice. Aim for 1 or fewer.

So a clear verdict is emerging: Ernest Hemingway was a poor to fair writer, whose writing is hard to read and who generally uses too many adverbs.

He's in good ("bad"?) company. The first two paragraphs of The Great Gatsby are rated "Bad" ("3 of 8 sentences are hard to read. 3 of 8 sentences are very hard to read. 6 adverbs. Aim for 1 or fewer.") The first paragraph of To the Lighthouse is rated "Bad" ("3 of 4 sentences are very hard to read. 5 adverbs. Aim for 1 or fewer. 2 uses of passive voice. Aim for 1 or fewer.") The first three sentences of the Declaration of Independence are rated "Bad"  ("3 of 3 sentences are very hard to read. 2 words or phrases can be simpler. 3 uses of passive voice. Aim for 1 or fewer.")

I could go on, but I think this is enough.

Update — It's worth noting that the "readability" index depends on word frequency (or maybe just word length?) as well as sentence length. Thus this document, though "Good":

is not quite as readable as this one:

Of course, neither one got dinged for too many passives or too many adverbs (since apparently to count as an adverb, a word in -ly needs to have another attested word as the pre-ly base…

And to understand this next one, you need to be in the 26th grade, which according to my calculations means four years of college (=16th grade), six years of grad school (=22nd grade), two years of post-doc (=24th grade), and then I'm not sure, two more years of kindergarten or something:

… OK, turns out it's just word length, not word frequency:



40 Comments

  1. Nathan said,

    February 13, 2014 @ 1:17 pm

    0 or fewer? Excuse my lack of confidence in this carefully-written app.

  2. Matt said,

    February 13, 2014 @ 1:18 pm

    Its built-in shock proof passive detector seems to be off, as "John was wugged," "John was red," and "John was Ted" all trigger a use of the passive voice.

  3. Darlene said,

    February 13, 2014 @ 1:36 pm

    Cool analysis. To be fair, though, I think this is more aimed at marketing/technical writers, where being clear and concise is much more important.

    It's a tool, not an end all be all sayer of what is good and bad.

  4. Matt_M said,

    February 13, 2014 @ 1:43 pm

    The app doesn't seem to understand syntax very well: I can count at least seven adverbs in the first Hemmingway passage, rather than the one spotted by the app (I wonder which one?):

    sure (in "sure never") / never/ only/ maybe/ early/ just/ just

  5. Jonathon Owen said,

    February 13, 2014 @ 1:54 pm

    Maybe the biggest problem is that they claim, "Hemingway makes your writing bold and clear." But the program has no way to evaluate what's bold or clear; all it does is flag long sentences, some adverbs, and passive constructions. Not only is it possible to flout all their advice and still write something bold and clear, but following their advice is no guarantee of good writing.

    For example, I just plugged in the following: "Hemingway am dumb. Program not knowing good writing. Least helpful advice ever?" Grade 7, good readability, no errors.

  6. Lazar said,

    February 13, 2014 @ 1:56 pm

    It seems to reflect pretty well the main superstitions of English writing: use simpler sentences, don't use adverbs, use shorter words, don't use passives. I've heard these things parroted by my teachers and professors, and – as this blog has documented – by almost all professional writers who claim to offer usage advice. It's clear that these dicta are almost completely divorced from reality, but what I still struggle to understand is how our society has come to such near-universal agreement on them. When and why did we all decide that intricacy, nuance and variety of form are so evil?

  7. Jonathon Owen said,

    February 13, 2014 @ 2:21 pm

    Matt_M: I just checked, and apparently the only thing it flagged as an adverb is "roly". It looks like it just flags pretty much anything ending in -ly, which would explain how it misses adverbs like "instead" in its own suggestions. I pasted in the first list from this site, and it thought 49 out of 61 of them were adverbs.

  8. Chris Waters said,

    February 13, 2014 @ 4:51 pm

    The example from For Whom the Bell Tolls appears to have a duplicated paragraph. I don't know if that's how you submitted it to the app, but it does seem to add some confusion (and a lack of terseness that Papa would never have allowed).

    [(myl) Oops, fixed now.]

    As someone in the industry, I can tell you that one of the scariest things is that some people who work at the company that made this probably actually believe in its power. Many years ago, I interviewed at a company that was making some program for analyzing business writing (not for clarity, but to try give you some sort of insight into the people who work at your company). At some point during the interview I said something about, "so it crunches the data and spits out some numbers," and the reaction I got was sheer shock. "It analyzes the writing" was the response I got. Needless to say, I didn't get the job, and was not unhappy! :)

  9. Rubrick said,

    February 13, 2014 @ 5:13 pm

    I'm a bit baffled by the commenters who seem to be trying to give this app some sort of benefit of the doubt. I think Mark demonstrated beyond a shred of doubt that it's completely and utterly useless. (Well, unless you count income derived from gullible app-buyers "useful".)

  10. A.D. said,

    February 13, 2014 @ 5:38 pm

    Input The Waste Land and received a grade of 3 with 0 sentences very hard to read.

    [(myl) Secret of his success… 3rd grade reading level, hardly any passives, no adverbs:

    Of course, "I was frightened" is not actually a passive…]

  11. Chris Waters said,

    February 13, 2014 @ 7:20 pm

    Really? I admit I'm not an expert, but "I was frightened" certainly looks like a passive to me. You can say, "I was frightened by X," and you can reverse that to get "X frightened me."

    Of course, as a regular reader, I'm not silly enough to believe there's anything wrong with a passive (clearly, the sled is the agent causing fright, and mentioning it a second time just to avoid the passive would be ridiculously redundant) but I'm still confused.

    (I do understand why "I was blue" or "I was stupid" aren't passives, which seems like it makes me smarter than this program, but that's not saying much!) :)

  12. Dave Harmon said,

    February 13, 2014 @ 8:35 pm

    Chris: "frightened" can also be used as an adjective, loosely synonymous with "afraid". ("The frightened birds scattered".) OK, it may technically be some complex form of a verb, but most people will interpret it as an adjective, and certainly that usage isn't actually passive.

  13. Luis Masanti said,

    February 13, 2014 @ 9:00 pm

    I won't defend the indefensible, but I'm with Darlene: This programs are not for literature.

    In the early ’70, there was the Writer's Workbench, a set of Unix programs to help you make more readable text… obviously, not literary texts.

    The usual joke was that if you put Shakespeare's work thru them, the results were awful.

    On the other hand, for programmers and engineers that usually are not the best writers in the world, those programs helped them to be clear, usually when writing manuals and instructions.

  14. Jon said,

    February 13, 2014 @ 9:08 pm

    Unfortunately, I was taught passive voice incorrectly. My teacher may have known better, but she "passively" reinforced the idea that any use of the verb "to be" is passive voice.

  15. Willi said,

    February 13, 2014 @ 9:09 pm

    I don't understand why this app rates adverbs as a bad thing. I'm European, and I've never heard anything bad about them in school. They were always equal to adjectives, if only more difficult to use. In fact I think adverbs are a rather pretty grammatical construction.

  16. Charles said,

    February 13, 2014 @ 9:48 pm

    I tried it on a blog article I wrote. I got an overall Good rating. It said my work was generally at 9th Grade reading levels, which is good because you should write for an audience below 10th Grade reading levels. Then it criticized me for having two sentences that were college level.

    I tried another article, a spoof that I deliberately made monotonous and repetitive. It complained that some of my sentences were college level, but half of my sentences were post-graduate level, including a sentence containing the phrase "the hell with this."

    Well the hell with that. I don't write for 9th graders. I write for me, and I have two college degrees. Henceforth, you must have a college degree to read my blog.

  17. Chris Waters said,

    February 13, 2014 @ 11:05 pm

    Dave Harmon: Duh, of course! Thanks. MYL was reading it one way and I was reading it the other, but on a second reading, I think MYL's interpretation makes more sense. Though they both work, which makes the whole thing charmingly ambiguous. :)

    Willi: as I said on FB when a friend asked the same thing: "Anything that can be overused (see Purple Prose) must be banned completely!" :)

  18. Brett said,

    February 14, 2014 @ 1:05 am

    The first couple of pages of the famously-unreadable Ulysses are apparently "Good" and Grade 5.

    I obviously should have given it to my son to read when he was 10.

    And just one "word or phrase can be simpler": 'however'.
    'Introibo ad altare Dei" on the other hand is just fine for readability.

  19. Jozef said,

    February 14, 2014 @ 3:57 am

    I still think the tool can be very useful, just not for fiction or long prose. Articles or reports targeted at a large audience, especially if they are non-native English speakers, should be short and concise. Using simple words is preferable. It is not a tool for professionals, but us – amateurs – can use it to write better.

  20. NW said,

    February 14, 2014 @ 4:50 am

    There was an Internet meme or thing several years back that claimed to analyse your writing and say 'You write like (famous author).' I put in a story of mine and apparently it was like Stephen King. Doubting this, I examined my vocabulary, changed 'wolves' to 'wombats' and made one other similar change into a less scary word, and lo and behold the grammatically and stylistically identical story was now like someone else. Huh. As useful as this Hemingway thing then.

  21. Lane said,

    February 14, 2014 @ 5:44 am

    Try "Buffalo buffalo Buffalo buffalo buffalo buffalo buffalo."

  22. richardelguru said,

    February 14, 2014 @ 7:11 am

    Jonathon deserves at least two Internets for his example in his first comment!
    Years ago in one of my silly radio essays I touched on the topic of on-line understandability analyzers (One Word in Four, but only if you're feeling particularly bored.)

  23. Dave Harmon said,

    February 14, 2014 @ 7:38 am

    Willi: Adding to Chris' sarcastic comment: Have you heard of Tom Swifties? (E.g., "'That dog's dangerous!', Tom growled menacingly.") Those have apparently traumatized a lot of English teachers!

  24. Off topic: Out-of-print books, Hemingway app, buy and draw, ‘LEGO Movie’ bloopers, sexy food | SiliconBeat said,

    February 14, 2014 @ 8:24 am

    […] books, according to BookFinder. Speaking of books, Language Log has some thoughts on the Hemingway app, which gives writing advice. Eight years is enough: A woman recently finished her project of […]

  25. John Saddington said,

    February 14, 2014 @ 9:32 am

    nothing but a solid LOL on this. Wow.

  26. Dmitry Mazin said,

    February 14, 2014 @ 10:56 am

    This app should have been called "Oglivy." Aka, write as if your readers are consumers looking at ads: simple.
    It teaches you to write like an adman, not a writer.

    [(myl) This is giving it way too much credit…]

  27. Charlei Bing said,

    February 14, 2014 @ 11:29 am

    The thing that bugged me immediately about this app's reports was the "X adverbs. Aim for 1 or fewer." I was taught that less should be used with things that cannot be counted and also with numbers when they are on their own. Should it be "X adverbs or less"?

  28. Jon Finch said,

    February 14, 2014 @ 11:39 am

    The question is not whether this app can help someone write good literature. Clearly it cannot.

    The real question is, where do they get the balls calling it "Hemingway"?

  29. Bill Benzon said,

    February 14, 2014 @ 12:02 pm

    Hemingway used his long sentences to do some interesting things. Here's a post where I argue that a sentence describing cape work in bull fighting is imitating the rhythms (in that cape work) that it describes:

    http://new-savanna.blogspot.com/2013/11/prose-rhythm-2-sentences-by-hemingway.html

  30. Grandpa said,

    February 14, 2014 @ 8:27 pm

    Jozef, you appear to be missing the point. It's not simply that this app does not appreciate great literature. The problem is that this app is literally junk. It is a con. It is not doing what it claims to do. It cannot analyse your writing in any meaningful sense. It does not know what an adverb is, or what the passive voice is; even if for some reason you want to avoid these things, this app cannot help you achieve. It does not know when a sentence is hard to understand; you may well want your writing to be easy for non-native speakers to read, but this app cannot help you achieve that. It is utterly useless for any purpose. It is not useful for anyone. Do not use it.

  31. Rick said,

    February 14, 2014 @ 8:53 pm

    @Charlei Bing @Nathan
    In response to the point about the use of "fewer" in the reports on adverbs, adverbs are clearly countable things. I can't see any problem with saying: "There are five adverbs in the passage. You should use fewer adverbs if possible." I know that "less" is gradually replacing "fewer" (just as "lay" is replacing the intransitive "lie") but I've never until now come across people trying to 'correct' "fewer" to "less", in the case of plural nouns—and I hope we all agree that "adverbs" is a plural noun—or indeed "lie" to "lay" when used intransitively.

  32. Nobody Important said,

    February 15, 2014 @ 2:32 am

    Any application telling me I have 1 items make me frown. But when the application is supposed to criticize my prose, telling me I used 1 adverbs makes me cringe. Double cringe for telling me to use 0 adverbs or fewer.

  33. Ray G said,

    February 15, 2014 @ 3:13 am

    I have to agree with Darlene and a few others who point out that this app is misnamed. Why would you use the name Hemingway when your app is more suited to very simple writing, like ads and some technical writing (like manuals, perhaps)? But even this is doubtful. I edit technical writing for a living, mostly articles destined for scientific and engineering journals. In those, flowery language (as in lots of adjectives and adverbs) is generally frowned upon, but the passive is widely embraced. And the use of adverbs or adjectives, per se, does not damn the writing as flowery.

    Regarding Lazar's comments on "the main superstitions of English writing: use simpler sentences, don't use adverbs, use shorter words, don't use passives." These guidelines probably arose among high school English teachers as a short, quick and dirty set of rules to help the otherwise hopelessly overmatched among us. If you've ever read some of what passes for written work in high school, and even university, you'd appreciate the need for such guidelines. They're not intended to guide you in your quest to write the next great American (Canadian, Australian, British, etc.) novel. They are intended to help you avoid looking like too much of an ass, or having your writing be totally incomprehensible.

  34. Chris Waters said,

    February 15, 2014 @ 5:55 pm

    Since people have been suggesting this was designed for technical writing, I gave it a try on some sample paragraphs from The C Programming Language by Brian W. Kernighan and Dennis M. Ritchie, which is widely considered a masterpiece of its kind: simple, clear, terse, easy-to-follow, and comprehensive. The results were, as I'm sure you'll be shocked to learn, about the same as Papa's: generally poor, with occasional moments that were just ok.

    K&R (as it's affectionately known in the industry) did score better on adverb use, which might mean something if this program were actually able to identify adverbs reliably, but since it can't, I don't think we can infer much there. :)

    I think the saddest part may be when it complained about a too-long sentence listing the elements found in C. I tried rewriting this as a series of short sentences. "C provides variables. C provides constants. C provides functions. C provides pointers. C provides arrays. […]" The app thought this was a great improvement! 'Nuff said! ;)

  35. Jay said,

    February 17, 2014 @ 4:17 pm

    You should give PaperRater.com a try. It's a similar app that does a much better job.

  36. Ray Dillinger said,

    February 17, 2014 @ 8:27 pm

    It would be interesting to create the framework of an application like this, backed with a stylometric analyzer, and then train it using a corpus of books widely considered to be good literature.

    For example, if you trained it on Hemingway's prose, then the stylometric analyzer would normally tell you whether a proposed passage is likely to have the same author, but the application as proposed would on that basis rate your writing on how much it resembles Hemingway's, possibly suggesting individual changes in words or phrases that would make the illusion more convincing (at least to the stylometric analyzer, which might care about things that completely escape human notice and be blind to things that do not).

    Such a tool might be useful if you wanted to produce convincing Sherlock Holmes stories, for example, in the style of Conan Doyle.

  37. Boris said,

    February 19, 2014 @ 1:58 pm

    @Ray Dillinger,
    Why stop there? Have separate corpora for "classical literature", "science fiction", "technical journals", "newspaper articles". This might actually teach you to write better. The only problem is, it probably wouldn't be able to tell you *why* a particular sentence is bad except "this word is not found in the corpus very frequently"

  38. Failed to understand the meaning of this "about" said,

    February 21, 2014 @ 12:18 pm

    […] knew so little of language that they had to fill in the blanks with these inanities. ////////// http://languagelog.ldc.upenn.edu/nll/?p=10416 It seems to reflect pretty well the main superstitions of English writing: use simpler sentences, […]

  39. Maggie McNeill said,

    April 2, 2014 @ 9:42 am

    Beautiful debunking of this monumentally dumb (yet highly praised) program! One of my readers posted a link to this column on today's edition of my own blog http://maggiemcneill.wordpress.com/2014/04/02/zero-intelligence/ which also criticized the program, though not as thoroughly as yours.

  40. Dan said,

    May 10, 2014 @ 10:02 am

    http://orwellapp.com is a similar idea to HeminwayApp but with a bit more functionality.

RSS feed for comments on this post