An experiment to test if Bionic Reading helps you read faster | Hacker News Reader

An experiment to test if Bionic Reading helps you read faster

blog.readwise.io

160 pointshoodwink93 comments

In my experience, more people need to work on reading comprehension than reading speed. Comprehension is what actually matters, and is what seems to be in short supply. If Bionic Reading promised increased reading speed with no loss of comprehension, and provided it, all it would mean is that people would misunderstand things at a faster rate.

I agree 100%. I think the fundamental flaw in seeking methods to speed read is the inherent assumption that there is an equal tradeoff between higher speed and lower comprehension. Alas, bad comprehension is probably many times more harmful than lower reading speed. So the tradeoff is highly imbalanced due to the highly dissimilar importance weights attached to comprehension and speed.

I’ve been thinking of that a lot as well. What’s the point of being able to read fast if you miss the fine details? What’s the point if you read so fast that you just sort of forget everything?

One of the many issues I have with contemporary literate culture, another being those who insist on reading a large volume of books, on being “well read”, rather than focusing on quality.

In defence of skimming books: Not everyone is enjoying reading the same way. Not even one person in different contexts. There are books where I'm going to be engaged completely. There are others I'll enjoy for main plot but will want to skip the fluff. I'm extreme cases I'll gloss over a whole paragraph as "whatever, I don't care about the scenery". Being able to skim quickly to the next interesting part is a good skill.

But the comment made me think - can we even improve the comprehension, as in processing what we read? (not as in understanding the vocabulary/structure) I've never heard of ways to purposefully do that.

For me an important part of reading fast is finding out which parts are interesting and which parts are not, and then focusing more on the interesting parts.

Reminds me of reading math textbooks in college. That text tends to be incredibly information dense, so I would spend minutes reading just one or two pages. One of my friends who tried to read them like they were any other book would get frustrated and stop.

The idea of reading at a given speed seems wonky to me as well. Its all about information density and how deeply you need to understand something imo. I read light fiction pretty quick, non fiction slower, technical stuff much slower etc

I find it really difficult and annoying -- and thereby slow -- to read with these "fixation" points as that's just not how I feel eyes work... I want my eyes to scan across the text rather than ratcheting from one word to the next word, which is both slow and exhausting. The words and "fixation" is thereby just a massive distraction. Hell: if I really really need to read quickly I try to do this thing where I scan my eyes diagonally downwards across the text to let my brain sort out entire lines of words at once... the last thing I want is to go single word by single word.

Personal anecdote. When I try to read fast, I notice fixation points as my eyes skip across the text. But it will be whole chunks of words. And anything that tries to draw my attention to one letter as opposed to another, will slow me down.

Note that my natural reading speed is ~950 wpm, so I tend to ignore these fads. I read fast enough already. The main barrier to reading faster is that it gets exhausting having my brain constantly trying to catch up to my eyes.

Note that my natural reading speed is ~950 wpm, so I tend to ignore these fads.

Were you just born with 4-5x the average reading speed? Or did you employ some other 'fads'? This seems pretty astonishing. Particularly in light of the following sentence on the "Speed Reading" entry on Wikipedia:

>Cognitive neuroscientist Stanislas Dehaene says that claims of reading up to 1,000 words per minute "must be viewed with skepticism"

I simply read a lot of science fiction/fantasy as a kid, and it happened naturally. I would view claims that a random person could be taught to read at my speed with skepticism. But I've met other people who read as fast as I do.

As a kid I thought I was simply a slightly fast reader until I happened to take a copy of https://www.amazon.com/Clan-Cave-Bear-Jean-Auel/dp/060961097... to a bath. I started, got caught up with the story, finished it, and finished my bath. My mother was so astounded that she quizzed me to verify I had actually read the book, and then estimated how what my reading speed had to be.

The weirdest thing is that my brain really does play catchup. I told my mother to test me by picking random spots and reading a few sentences. I was then able to tell her what was going on, and when handed the book could find the passage. That I was able to do. But the book was still jumbled up in my brain - I couldn't have given a plot summary for a day or so.

I strongly suspect that this kind of "brain pipelining" with large amounts of buffering is critical for really fast reading. Get every slow step out of the loop, only do what's fast. Let the slow bits of your brain catch up.

Does that take any enjoyment out of reading for you? I tried speed reading for a brief stint, but then decided to stop. For me reading a good book is like savoring good food, I wouldn't want to chug it down.

Not speaking for btilly here, strictly for myself: I think it depends on the kind of book, kind of writing, and personal expectations. When reading a novel, I'm mostly interested in the story and character development, and less in things like long descriptions of the environment (which Jean Auel does, a lot, though maybe not as much in the first one IIRC). I'm like the kid always asking "what's next" even before the narrator finishes his sentence.

So I like to read fast, and maybe there are some things I'm missing but I don't feel like that's a loss.

Mind you I'm not really speed reading (I think), I just read pretty fast. And skip uninteresting pieces of text (by skip I mean read some words or parts of sentences here and there, like the start and/or end of paragraphs, to know what the text is about so I know where to resume normal reading).

Now food is something else entirely :) I'm a pretty fast reader but a slow eater.

It depends on the book. Most sci-fi/fantasy I enjoy at speed. Anything where I want to find out what happens NEXT. But poetry and math I would want to read more slowly. I read more quickly than I think, and a full appreciation of the rhythm of language takes time.

What I do isn't what most people call speed reading, though. I don't skip any words, and can recognize and pull out exact phrases. I don't know how much text I could hold like that. When I was 19, a long book. But I think my capacity has gone down with age.

Tangential personal anecdote.

I am autistic and I have two unusual informational input traits.

I would consider myself a good reader. A sizeable chunk of my life has been dedicated to reading (and absorbing other forms of media). I can't read at a 950 wpm (that figure is immediately raising suspicions for me) but I'm still fairly fast.

The first weird thing:

I taught myself how to read and I've been reading since I was 3 years old. Apparently this freaked out my aunt when I was in her car reading street signs aloud. I still have memories of being far ahead of my peers in early childhood. School was unable to challenge me and this led to me having a lax study attitude and I became lazy. As an adult I'm still lazy, but I've been able to turn this into a strength as a programmer. (See Bill Gate's quote on "a lazy person").

For anyone else who has or knows someone who is experiencing this: HealthyGamerGG's video: "Why Gifted Kids Are Actually Special Needs" can give some great information to help understand this.

The second weird thing:

I regularly watch informational/tutorial/conference/etc... videos on youtube between 2-4 times their standard speed. I do this in the browser's console with the following command:

$('video').playbackRate = x;

Where x is a number. (e.g. 1, 3, 2.75, etc...)

After you've already typed it once, a simple press of the Up key will bring it back as if you had just typed it.

I've been told there are extensions that do this while avoiding the terminal, but this is already ingrained in my muscle memory. (F12 -> Up -> Delete -> type number -> Enter -> F12.)

Understanding sped-up talking is a skill I've built up over time. To other people around me who have tried to follow along it sounds like gibberish. I've heard of deaf developers who commonly develop this skill so that they're listening 600-800%+ standard speed but I don't think the upper range is possible on videos with different voices and accents using a wider vocabulary.

I'm like you in both respects, but I also read as fast as the parent commenter. My operating theory had been that, since nobody had taught me to read one word at a time and subvocalize them, I just developed a different means of reading than most people, reading several lines at a time rather than individual words.

The jury is out as far as whether I'm also autistic, though. I have a great many traits in common with people who are, but I've also found a surprising amount of success reducing my more frustrating symptoms by a combination of working through early childhood trauma (which has helped my nervous system stop overreacting to stimuli) and extremely strict regulation of my diet (I have only eaten 10 foods, prepared in two ways, in the last year and a half) to reduce gut leakage and brain inflammation. (The diet is not expected to be this severe forever. Foods are reintroduced in stages. And it includes slow-cooked meat on bones and organ meats, so it's not as dire/risky a nutritional thing as it probably sounds.)

Because I had no language delay as a child and Asperger's was not a separate diagnosis until I was well into elementary school (and they certainly weren't looking for it in girls), I was never tested or diagnosed as a child. I also haven't identified enough advantages to pursuing a diagnosis as an adult to choose to pursue it now. But if healing my nervous system and my gut eliminates my symptoms (eventually), I'm inclined to think that would indicate I didn't have autism in the first place, just a cluster of overlapping symptoms.

950 wpm is impressive.

I did the test from the article, and got about 500 wpm with the standard font (about 400 wpm on the Bionic Reading thing). Even that speed is at the limit; I'm already skipping parts of the text that don't seem to contribute much (I could still correctly answer the reading comprehension quizzes at the end though).

I ignore all those fads too. This Bionic Reading font in particular feels completely backwards: it tries to make me read word by word which is needlessly slow.

> I want my eyes to scan across the text rather than ratcheting from one word to the next word, which is both slow and exhausting.

An interesting experiment is to try reading a word at a time, but without moving the eyes. Here's [1] a command line program that takes text on standard input and displays it one word at a time in a fixed position on the screen, holding each word for N milliseconds (3N milliseconds if there is punctuation) where N defaults to 250 but can be set on the command line.

At the default 250 msec per word I found it very easy to read the material I tested with (random extracts from a Project Gutenberg edition of"The Valley of Fear", by Arthur Conan Doyle). That works out to around 160-180 wpm. (Not the 240 wpm you would expect from 250 msec/word because of the delays for punctuation).

At 200 msec/word, it still feels like a very slow read. Rate was around 270 wpm.

150 msec/word gave around 320 wpm. Still not a problem keeping up.

120 msec/word pushed it up to around 380 wpm and it starts to get hard for me. If I don't quite catch a word and have to think a little to figure out what it was I can get distracted enough to miss more words unless some punctuation comes up soon to give me a little break.

100 msec/word, around 480 wpm, is still reasonably comprehensible but at that point requires a lot of focus and feels tiring even though my eyes don't have to move. Sometimes not moving can be as tiring as moving when you are trying to not move for a long time.

I would not want to read a lot this way, but there are some places I wish it were offered. Many music players for example if the title does not fit in the space available autoscroll it back and forth. It can be very hard to read it while it is scrolling. A word flash display might work better there.

[1] https://pastebin.com/70eYTDkk

I saw that demonstrated on Hacker News some years ago, in a web form.

It. Was. Horrible. For. Me.

My eyes are used to processing much bigger chunks than a single word, and know how to move to the chunk size that they used. Therefore I topped out at a fraction of my usual reading speed.

The evidence regarding comprehension using RSVP is similar to what TFA mentioned with Bionic, which was that comprehension drops as speed goes up.

The last company that tried to popularize this was Spritz, and I believe they are now owned/operated by their erstwhile VC. It probably didn't help that the proprietary bits they added were not compelling, so other companies just implemented vanilla versions and therefore didn't have to pay them anything.

I got this feeling as well and I did 50% better with the plain Literata. I wonder the bolding might speed up slower readers and slow down faster readers by encouraging a word-by-word pace. It would be interesting to see a scatter plot of Bionic vs Literata speed.

Yes, I think some of the popularity with Bionic reading is for ADHD users who I believe are slow readers in part because they have to go back and reread because they lose focus. I am one of them, and according to this, I am 15% faster with Bionic reading.

It's cool to see that someone is trying to test these claims. Not long ago, the BR site was shared on my work Slack and there were some that immediately argued "Yup, I can read faster using this and comprehend everything", which is quite a bold claim to make after reading a paragraph. Personally, having bolded letters sprinkled across my screen feels noisy and distracting, so it's a bit hard to get past that mental block.

https://speed.readwise.io

   Your results

    Typeface         Bionic Reading     Literata

    Article 1        591 WPM            352 WPM
    Article 2        657 WPM            457 WPM
    Average Speed    624 WPM            405 WPM

    You read 43% faster with Bionic Reading

Guess gotta give this a try for storybooks...

maybe some enterprising and bored HN nerd can make a Calibre plugin that converts regular epubs to bionic reading enabled epub files

----

but the truth is, for the aforementioned storybooks, often I just LOOK at a whole paragraph or even the whole page and just pick out the relevant word of two in a story.

Unlike actual educational content, where the exact text matters, in fiction, after having developed a hobby to read for entertainment for so long and having gotten used to so many tropes, I just often just breeze through and look for the word that confirms which direction the author is taking that paragraph in, and often just glide over the paragraph and go to the next one.

It's like learning how to drive, I guess. At some point you are not supposed to look at every thing, you just take the overall picture and just go.

Not sure what is achieved by making every road sign and billboard flashing neon is supposed to achieve in such a scenario. Not all words are worthy of equal attention, most are meant to be glossed over.

____

but for actual educational purposes, the best test IMHO is to use it for the boring but important texts, like training manuals. See if it actually helps people learn and retain more import information about the new tool or procedure they are learning about.

This technique will work for most text, but can be deliberately hacked in a bad way. I discovered this about 20 years ago from a cleverly written attack post on Perlmonks. Every paragraph started with a positive sentence whose tone slid into negative by the end of the paragraph. And then to positive for the next paragraph.

I was horrified at how an attack piece had so many positive votes, and asked people why they were voting for it. It turns out that a lot (probably most) had just skimmed it, felt good, voted for it, and moved on. They literally hadn't seen the nasty things said about a variety of people.

That's why I specified story books instead of online articles, because with story books, you build a rapport with the author and you have kinda narrowed down the options they will take, on a meta level.

But good point on hacking people's tendency to glide, I shall keep that in mind.

Jeeze that is fast. More than twice the average reading speed. I wonder if bionic reading helps more for "speed readers"? Your results surely suggest it.

Also - It must be so painful for you to type, or heck even speak!

I dunno if I'm a "speed" reader, or just a fast glosser.

Also, I am not quite sure if I really understood the text... or used my knowledge of the views of the "tech" community, and just "guessed" which of the three mcq options paul graham probably meant?

It could be I just bullshitted my way into correct answers...Honestly, after a night's sleep, I don't recall much of what was said in those articles, so how much did I actually retain?

That said, I do find it painful to watch any content on less than 2x speed, but that maybe because I'm probably on the spectrum.... Things get complicated.

A few pages down is the answer to what (hopefully) many are wondering:

"Here's how we've designed the pilot study:

...

4. After finishing each document, we asked three multiple choice questions to control for and confirm comprehension."

Here are some bookmarklets I developed to cut out the extra formatting, bolding, and other content on websites that makes them difficult to read. I would appreciate any feedback or modifications for improvement.

Literata font with a peach background color to optimize reading speed and no selective bolding:

javascript:void function(){javascript:(function(){var a=Math.floor,b=document.querySelectorAll("p, title, a, ul"),c=[],e="",f="",g="",h=0,k=0,l="",m="",n=window.open("","_blank");for(var d in b){var i=b[d].textContent;i%26%26(c=c+"\n"+i)}for(f=c,e=f.replace(/\n/g," "),g=e.split(" "),h=0;h<g.length;h++)k=a(g[h].length/3)+1,l=""+g[h].substring(0,k)+""+g[h].substring(k,g[h].length)+" ","."==g[h].substring(g[h].length-1,g[h].length)%26%26(l+=" "),m+=l;n.document.write("<html>"+m+"</html>")})()}();

Selective bolding with a peach background for those who find a benefit:

javascript:void function(){javascript:(function(){var a=Math.floor,b=document.querySelectorAll("p, title, a, ul"),c=[],e="",f="",g="",h=0,k=0,l="",m="",n=window.open("","_blank");for(var d in b){var i=b[d].textContent;i%26%26(c=c+"\n"+i)}for(f=c,e=f.replace(/\n/g," "),g=e.split(" "),h=0;h<g.length;h++)k=a(g[h].length/3)+1,l=""+g[h].substring(0,k)+""+g[h].substring(k,g[h].length)+" ","."==g[h].substring(g[h].length-1,g[h].length)%26%26(l+=" * "),m+=l;n.document.write("<html>"+m+"</html>")})()}();

Probably a good idea to add `font-size: 0.5cm` in there or something.

Did you never find the Reader Mode button?

Linked at the end of the article is the short (15 minutes maybe) self-test: https://speed.readwise.io/.

Curious what results people here got. I got 23% faster with BR but I couldn't tell if I was just trying harder.

23% faster with BR as well. 528 vs 419 WPM. Made a point about feeling comfortable while reading.

I was a mere 2% faster. 404 vs 396 WPM.

Edit: 2% faster with BR.

we'll definitely make a follow up post once we've gathered enough data here! hopefully in a week or so :)

(my results were 2% faster with BR)

have you considered implementing a false control (i.e. putting in a badly implemented bionic reading version and serving that to a random number of people to see whether people also report speed increases with "bad bionic reading"?)

i know a significant number of people here are probably already familiar with what bionic reading is, but still, could be an interesting control.

yeah i acknowledged that in the experiment design discussion

> We recognize this is clearly not a gold standard, double-blind RCT. Researcher and participant alike knew what Bionic Reading claims to achieve. We simply did not have time to (1) develop a mock Bionic Reading placebo that, for example, might bold the middle of the word instead of the beginning and/or (2) find a set of test subjects who are blissfully offline and unaware of this viral sensation. That said, we controlled for what we could and we think we've already gathered some fairly compelling evidence.

I got 15% faster with the non-BR font. I was trying to do the same amount on each.

I got 1% faster with Literata (385 vs 390). Bionic made my eyes stutter so much I just didn't like reading it either.

12% faster with Literata.

I was 15% faster with Bionic reading, but I noticed that my comprehension suffered somewhat. I still got all the review questions right, but I sometimes had to re-read sentences or phrases and didn't have that problem as much with the normal text.

My assumption is that I'm able to scan the words better with BR, but my eye gets pulled along too fast and I'm reading without comprehending.

I was 17% faster with Literata.

I can't tell whether I was intentionally going faster because I had preconceived negative opinions about BR. It's a possibility.

15% faster with Literata (381 WPM with BR vs 442 WPM with Literata). Not sure if that's accurate since I'm a bit impatient at the BR part.

18% faster with Literata

Also 23% faster with BR but I enjoyed reading that way less. As with yourself, I suspect that maybe I put pressure on myself to read faster for the BR.

As the article says, a better comparison may be to have a fake version of BR as the control.

It's not a perfect test, but as a statistician, my preliminary takeaway from this is that bionic reading doesn't work.

results are generally in the ballpark of insignificant, and there seems to be a very weak inverse relationship between reading speed and comprehension.

which is to say, on preliminary it looks like most "speed reading" claims (I.e. quackery).

I read once that when we read, our eyes bounce along the tops of the letters. A demonstration removed the lower parts of the letters, and interestingly, I was able to still read the words. I would like to see this same test but with a font that removed the lower half of the words in each sentence. Just for kicks and giggles...

I have dyslexia, and I can't scan read(if that is the right word), I have to basically read entire words. This stuff forces me to scan read, which means I just don't understand the text anymore.

If the authors of the article are reading this: I would also do a test to see if the variance of the two groups (bionic vs normal) is significantly different. I did not check it but bionic might be very beneficial to some and detrimental to others which would not be seen when looking at the mean (you should also plot the distributions, it pretty much always a good first reflex).

for sure. due to this improved experiment design, we should now have the data to perform paired analyses rather than just looking at means.

what test do you use to compare variances though?

There are several options but the Levene test should be appropriate: https://www.itl.nist.gov/div898/handbook/eda/section3/eda35a...

I notice that I no longer spend energy actively scanning lines with bionic reading - my eyes snap automatically. I'm able to look at the text "from a distance" and the words just enter my mind, and I did get 31% faster reading from bionics according to this test.

A whopping 0% here ><

272 vs 272 wpm, English is not my native language and I just ran a 5k, not sure if that has any relevance

i could concentrate myself better on the regular text

with Bionic Reading i had to re-read some sentences until i've processed them correctly

note that i'm not a native speaker

personal result:

Bionic Reading: 234 WPM

Literata: 262 WPM

"You read 11% faster with Literata"

> When you click the button below, you'll begin reading the first of two Paul Graham essays, each approximately 5 minutes in length.

My idea of hell. I'll wait for a better choice of source material.

>you should notice a funny pattern: almost every person's fastest font was the font they received on the first article.

Is it possible that people "calibrate" their brains towards a specific reading style, and that first style induces the calibration? Similar to how you can calibrate to quickly pick out a picture of a specific object, from a gallery of pictures, after you've prepared yourself to see that picture.

I'm guessing that the very next sentence was the actual reason:

>This is likely because the first article was less challenging to read

That doesn't account for why the font affinity carries over to the rest of the test, which is why they propose multiple theories, but aren't sure.

Faster with Bionic by 6%. Less of a difference than I expected although it does feel easier for some reason.

I have thought about using Bionic or Dyslexic fonts but I'm worried that it'd cause a change over time that makes it even harder to read plain format text.

My reading speed overall is slow compared to others. I notice my reading speed is only fast and accurate when I'm extremely interested in a topic.

I wouldn't be too concerned that using specialized fonts would become a dependency. I run BeeLine Reader, a similar tool that uses color gradients instead of bolding, and we've not heard this from our users much (though we did wonder about it when we were starting out). People do like specialized fonts and wish they had them on platforms where they're unavailable, but it's not like it becomes a crutch that you can't live without.

778 WPM with Bionic, 672 without. Bionic appears to have increased my speed by 15% but it's also possible that I was just paying more attention to the articles implementing the technique. It's unfortunate that the comprehension scores aren't given though (unless I'm missing something?)

sorry about that. that was an oversight on our part, but we're collecting the comprehension scores in the dataset and will include in our results hopefully sometime next week.

Bionic reading: 317 WPM Literata: 317 WPM You read 0% faster with Bionic Reading

I guess I’m a pretty consistent slow reader.

Way faster than me, and at the top end of average, based on everything I can find online. I'm curious though, what the average is among "readers" lol.

Bionic 633 wpm, literata 803.

I do read quickly but I have read that essay before, so perhaps I only skimmed it. I wasn't certain on one of the comprehension questions.

421 wpm with bionic reading, 447 without, but I had the impression of being faster with bionic reading, something about reading being more smooth. I guess it's smoother but slower overall.

nice! Surprised to see I ended up reading faster without bionic reading

I don't understand.

From their own metrics - speed went up by a lower percentage than comprehension dropped.

This sounds like a terrible idea...

I don't think they are comparable..

I’m not 100% sure it works for everybody, but it probably helps that English have shorter words where the stress tends to be on the first syllable, doesn’t have genders, always requires an explicit subject and has simple verb conjugations.

Would it work with languages that have longer words and a case system?

19% faster with Bionic Reading. I'd like to try again with different source articles that are not already part of my field of expertise. I felt as if I could answer the comprehension questions without having read the articles.

But does it help you retain what you read? Speed workout comprehension is pointless.

I take it you didn't read the article? It answers that (ish)

Thanks. I was busy and have seen several naive articles promoting speed readying recently and assumed this was one more.

Interesting website. I never considered in my 30-something years of reading that I was a slow reader but apparently I am (<250 WPM). Most of my reading (outside of the internet) is scholarly articles for my job.

10% faster with literata, altho with bionic i had 0 issues losing tracks of where I was or skipping lines but accident, which made me think i was reading faster, but apparently i was not.

Many of the comprehension questions could have been answered without reading the article. But some were quizzing irrelevant parts of the text. It's a hard balance to strike

Your results Readwise logo Typeface Bionic Reading Literata Article 1 479 WPM 574 WPM Article 2 463 WPM 507 WPM Average Speed 471 WPM 541 WPM

You read 14% faster with Literata

I get the opposite effect. Instead of reading each word, I start reading the parts of the word which is not normally how I read. It's very jarring.

Same. I found it impossible to concentrate. Similar disrupting effect as gifs (visual noise) next to the text or similar.

I found Bionic 10% slower than Literata.

But I also found I am a verrrry slow reader. My reading speed came to 147 wpm. Any recommendation to improve this?

When I was in school reading coaches encouraged the awareness of subvocalization (where you mimic the text as if you were speaking it, but not aloud). If you find yourself doing that while reading, stop/suppress that.

Why do people want to read faster? You can also write tersely. Or not overload yourself with useless information

1% faster with bionic reading, but I was eating chips and dip so maybe I was just distracted by stuffing my face.

Another sort of physically interesting thing that might unexpectedly help some people:

read with the opendyslexic font.

I believe the screen is a huge factor in it as well that should be accounted for.

we’re capturing desktop versus mobile :)

respect author do himself experiment! he even calculated p-value for it, also result did not recommended his method, but he let me trust he in future.

Hit finished reading and got a blank page…

Was there an ad or something?

There wasn't. What lists is your adblocker running? I got my results with standard uBlock.

Default list for 1Blocker

400ish wpm with normal text, 0wpm with "Bionic Reading" wibbly wobbly fonts. It just doesn't look like text.

This is really cool

Reply on news.ycombinator.com