We found no evidence that Bionic Reading has a positive effect on reading speed
blog.readwise.io
blog.readwise.io
So sure reading fast would be nice, but I think that’s mostly biological whereas slow quality reading you have full control of and we need to focus on that instead.
Ironically (or not?) self-help books are absolutely full of fluff but at their core can have very useful/helpful ideas. They just get expounded upon back up with 10 specific examples where each example starts at the beginning of the person's life in full detail.
Speed reading has helped me get through those books and learn something where I would have wasted so much time or actually never even picked the book up.
That's not fluff. It's an artifact of people being different and the goal of the book being to connect with someone and make an impact. A list of pithy principles would have no impact and make no connection. The exposition can do that (e.g. relatable story, example of an application that's close enough to the reader's circumstances), but not all of them will connect with every person.
In case anyone wants to put it to the test, here is a book consisting of precisely that (by Rochefoucauld, from the 1600s): https://www.gutenberg.org/files/9105/9105-h/9105-h.htm#linkm...
It then seems to me these kinds of books must be made for people who want others to think for them.
So maybe all these books should say is: "Think. Your problem is a lack of aforementioned activity. If you need some food for thought here's a list of ten pithy principles. Flip to page two for an afterword by my publisher."
Of course that wouldn't be good business sense. Why show someone the spring when you can sell them water?
He can't. The author can't predict which story or example will connect with you, so he includes many that connected with someone, with the hope that some fraction will connect with any given person. That's why people complain about "fluff": they're annoyed but the stuff that doesn't connect with them, but they haven't thought about it beyond their own personal experience. Maybe they could redact the book down to 10 pages, but so can everyone else, and all the redactions would be different.
> It then seems to me these kinds of books must be made for people who want others to think for them.
You're being uncharitable and kind of conceited. Would you say Calculus textbooks are for people who want others to think for them? Do Real Men take a short primer on mathematical logic and the axioms of ZFC set theory, and go derive Calculus for themselves?
Emphasis mine.
> Would you say Calculus textbooks are for people who want others to think for them? Do Real Men take a short primer on mathematical logic and the axioms of ZFC set theory, and go derive Calculus for themselves?
My response was argumentum ad absurdum, so I gain nothing from defending the position, but still:
Walter Rudin's Principles Of Mathematical Analysis would be a terrible book if it tried to relate the matter's purpose at every step. Nobody expects a mathematical textbook to do that.
In fact Principles Of Mathematical Analysis is a great book for being extremely concise and containing just what is necessary for a reasonably intelligent reader to understand the material.
To be more clear, I don't believe that your explanation could possibly be a good reason for these stories being included in a self help book, but without precluding they may serve another purpose.
What about readers who aren't "reasonably intelligent" (which often means "quite a bit more intelligent than the typical person")?
> To be more clear, I don't believe that your explanation could possibly be a good reason for these stories being included in a self help book, but without precluding they may serve another purpose.
What's the reason for your emphasis there? Are you reading the "self" part too literally or idiosyncratically? IIRC, the "self" just means the book is meant to help the reader with his problem without personal guidance from some professional. It doesn't mean the reader is supposed to figure it out on his own.
As I've gotten older, I've gotten more wary of certain biases that engineer-types often tend to indulge in. One of them is along the lines of "I'm so smart, I think I can figure it out on my own, therefore everything I think I don't need is unnecessary." Another is temptation to confirm one's intelligence by seeing the "real" reason as some cynical ploy that works on lesser people.
Also, I'm not saying every self-help book is good, or that it never happens that examples are truly just padding. It's just that there's good, non-cynical reasons to not to reduce everything down to some pithy list of axioms, and I know for a fact that at least one well-regarded one is structured that way, and it was a bit of a slog because of all the examples and stuff that didn't connect with my particular circumstance (but there's no way for someone I never met writing a couple years before my birth to tailor anything to me).
Examples are shortcuts find relevance. Once I’ve decided a topic is relevant to me then I think deeply about it.
It's been increasingly pointed out to me that my fiction reading habits, which permit enjoyful re-re-re-reading (the constant surprise of things missed each prior reading) are anti-patterns for legal or technical writing: in these cases reading deep, slow, is far better.
There's nothing wrong with re-reading per se. It's about context.
Not everything should be enjoyed though and that needs to be better known lest people feel they’re doing it wrong when reading skimmable material back to back.
https://www.tsw.it/wp-content/uploads/Rapid-serial-visual-pr...
One of the issues with tests like these is that the companies sponsor the research and they are one-off vs long exposure studies; that is as with most interfaces with high-throughput, takes time for the mind to adjust.
My guess that given human speech appears to have a universal transmission rate of around 39-bits per second, that on average that’s going to be the actual performance target:
This is unacceptable, we need a modern language that packs information tighter!
Discussed in HN multiple times https://news.ycombinator.com/item?id=18703805
Evidence suggests the real limit is how quickly human brains can take ideas/qualia, convert them into abstractions, and encode the abstractions into language. This is because (A) very different languages still exhibit similar limits and (B) those limits appear to be governed by the sending-side. People can comprehend spoken words at a higher rate than they can spontaneously emit them.
So trying to make the language more "compact" would likely just waste precious brain-cycles on the compression step, which isn't actually necessary when your mouth already supports talking faster.
Technology analogy: Two computers are collaboratively solving a problem with back-and-forth messages. The network connection is actually very good, however the bottleneck is the CPU in each computer. Will the problem be solved faster if you change the transmission style from plaintext to gzipped?
If you have any links to research supporting this, I would be interested.
> The mean Listening Rate was 56.8, which corresponds to 309 WPM. Given that people typically speak at a rate of 120-180 WPM, these results suggest that many people, if not most, can understand speech signifcantly [sic] faster than today’s conversational agents with typical human speaking rates.
While there was some difference between the visually-impaired and normal-vision respondents, I don't think it's enough to matter to the thesis of my post:
> [...] The mean Listening Rate for visually impaired participants was 60.6 (334 WPM) while for sighted participants it was 55.1 (297 WPM).
From page 4 of 12 of the PDF you linked to, “Rhyme test: measures word recognition by playing a single recorded word, and asking the participant to identify it from a list of six rhyming options (e.g., went, sent, bent, dent, tent, rent). We used 50 sets of rhyming words (300 words total), taken from the Modifed Rhyme Test [27], a standard test used to evaluate auditory comprehension.” The word list research used is on page 30-31 of 55 of this PDF:
https://www.researchgate.net/profile/Michael-Hecker-3/public...
If true, test appears to not even measuring word recognition, it’s more accurately measuring a single phoneme recognition. If the listener correct picks the correct phoneme from a multi-choice list, researchers assume person would hear and understand 100% of any expressions received at that rate of speed; which in my opinion is clearly flawed.
I would be the first to agree that testing listening comprehension rates is hard to do, hence why I asked to review research, but to me, unless I am misunderstanding something, unclear how this research actually provides any meaningful observations.
That being said, this has been attempted before: see https://en.wikipedia.org/wiki/Speedtalk and https://www.zompist.com/kitlong.html#howmany, and especially https://web.archive.org/web/20000503004430/http://fatmac.ee.... for a detailed attempt.
Like so many things with people, it depends on what they practice. Most people practice listening far more than rapid speech. Some people can speak faster than most people can comprehend. Part of learning to do that may involve not listening (to themselves) in the same way.
If you learned this new more compact language as your native language or to native fluency, I don't imagine you would need to go through a compression step, since all of your thoughts would already have happened in the compacted form.
One option is that what you're describing just isn't possible, that humans are already butting up against some kind of limit which is not avoidable simply by being raised with a different language.
Another option is that one can be raised to think in "pre-compressed" structures, but nobody does because it's a bad tradeoff, dropping "general thinking" performance with a worse impact than any "faster speaking" benefits. (Such as being simply slower, or more error-prone, or more demanding on attentional resources, etc.)
Maybe reading a computer screen is a simpler task than talking person to person, but it's an interesting datapoint!
On the flip side, I provided research on transmission rates, which to me seems reasonable, but another user shared research on reception rates, which to me is unreasonable:
https://news.ycombinator.com/item?id=32160095
To me, I am interested in notable, reviewable progress in understanding the topic — not chatting about it around an internet campfire.
The authors then average results (good and bad) across all users, resulting in a number close to zero, and conclude Bionic Reading doesn’t work for anyone, even calling it a placebo effect at the end.
The problem is, all brains are not the same. It doesn’t have to work equally well for everyone to be valuable.
Well, that's not really something you'd want to admit in your paper, much less advance as an argument. The obvious conclusion is that there's a mistake in your data, not that some users will see an increase in reading speed of 4.9 words per second.
You can make the _claim_ in your comment, but there isn't statistical evidence that this effect is more than random, or if the same people were tested again their results wouldn't be opposite. Put another way, it doesn't pass the null hypothesis.
> Since posting this experiment, I've received a lot of side comments along the lines of, "Well, of course I don't expect Bionic Reading to work for most people, but for [my subpopulation], it really works." If that were the case, we might expect to see disproportionate benefits for those participants who read faster with Bionic Reading than for those who read faster without Bionic Reading. Let's look at how many participants read faster with each font and their average speed gains.
Table 3: Summary speed differences per faster font Count Percent Delta (WPM) Bionic 998 52% 35 Non-Bionic 918 48% 43
> The number of people who read faster with Bionic Reading was slightly greater (52%) than the number of people who read faster without Bionic Reading (48%). That said, those who read faster with Bionic Reading only picked up 35 words per minute on average. In contrast, those who read faster without Bionic Reading picked up 43 words per minute. It does not appear that when Bionic Reading works, it really works.
I agree though, it would have been an interesting follow-up if they asked only people who felt they benefit from it, and then conducted their analysis again on that sample!
Your point is especially apt here, where there is a possibility that the technology could be helpful for a subpopulation, and act somewhat like an assistive technology. It is very tricky to assess the utility of assistive technologies (I work in this field and have asked many experts how they do it) because the average impact isn't the most important thing. It doesn't matter if high-contrast mode is bad for 90% of people, if for the 10% who would actually use it, it's helpful.
One way that the study could have been optimized is if participants had been asked (after taking the test, to avoid priming) if they have any specific reading challenges. That would have helped identify whether there appear to be any subgroups that disproportionately benefit from the approach.
1: https://www.amazon.com/End-Average-Succeed-Values-Sameness/d...
Why do we only crunch data down to averages? Differences in group variation can be measured just as rigorously as differences in group averages. Repeatedly testing an unusual individual is perfectly legitimate and scientifically interesting. There is no Law of Science that says you have to execute a boring protocol that rigidly assumes everyone is the same.
There's a classic story about how the Air Force learned that no one is average, and switched from fixed "average" cockpit fittings to adjustable ones, greatly reducing accidents. But it seems like no one has learned from this.
For heaven's sake, let's try to learn about individuals and what works for them!
BeeLine founder here — feel free to reach out prior to your testing (nick@[domain]). We can share info on past testing and implementation techniques.
We've wondered about Sans Forgetica in the past, which may increase reading comprehension (at the expense of speed). It might be interesting to try a BeeLine/Forgetica test, to try to get the best of both worlds!
sed 's/ and / \& /g' -e 's/ of /\ % /g' -e 's/ is / \= /g' -e 's/ th / \# /g' -e 's/\B[aeiouAEIOU]//g' -e 's/\(.\)\1/\1/g' -e "s/'//g" sacred-world.txt > sw.txt
I'll be hanging out in the replies today if there are any questions I can answer :)
Is there a way for me to buy you a coffee (or some cash equivalent)? I'd love to incentivize such rigor in online discourse.
An obvious way is to try out readwise ;) but I mostly read online articles and (pirated) books, so I'm not sure I'm a good use case.
we've actually built an app for reading articles and ebooks (and RSS and PDFs and email newsletters and Twitter threads): https://readwise.io/read
still in private beta but we'll be entering public beta before summer is over!
I'd like to see what the difference would be like after a day, four, and a week of reading articles with BR, if that's viable to enlist people for at all.
And it turns out that comprehension is suffering from trying to read faster... I think that most people, when they read, do a sort of automatic "speed reading" in their mind already.
What I mean is: I notice that when I read, my eyes will skip parts of words where I don't need to see all the letters to know what word it is. What if all this trickery simply hinders our innate and built-in "automatic speed reading" capability?
If it were a form of speed reading, we would have observed faster speeds when using Bionic Reading (and probably lower comprehension as a trade-off). This is NOT what we observed.
Not a summary, but kind of like a lossless re-encoding for maximum human efficiency.
You could even allow custom reader profiles that take into account things like vocabulary an subject expertise.
[0] https://www.reddit.com/r/autotldr/comments/31b9fm/faq_autotl...
But for people who want to read faster in a lossless way this wouldn’t apply.
Unrelated Readwise questions (love the work y’all do, thank you!) - any idea when the chrome app will support multiple highlights in one go? - is there a random quote API that references all sources at once? - has the team considered adding semantic search to the app?
An AI could also improve on the status quo by being more consistent and there's _some_ possibility that it could be less biased than in human in several ways.
on AI-based booked summarization, here's a really interesting article by OpenAI on the topic: https://openai.com/blog/summarizing-books/
*any idea when the chrome app will support multiple highlights in one go?* we have a new browser extension (yet unreleased) that enables you to highlight the native page (i think that's what you're asking)
*is there a random quote API that references all sources at once?* at the individual user level, yes. readwise.io/api_deets
*has the team considered adding semantic search to the app?* not yet, but my cofounder tristan did push a huge full-text search update a few months ago that makes search results 10x better than they were before
You get better at reading by reading and deliberately practicing it. Speed reading is a completely different skill than comprehension. Comprehension is a completely different skill than entertaining yourself. The fundamentals aren't going anywhere. They'll be here when you realize speedreading, bionic reading, and summaries are just distractions.
There would be no difference between a great teacher and a terrible teacher otherwise. Or a great coach and a terrible coach.
I don't know if this is because the subject matter is novel, or that the authors are better at math and science than writing.
I wonder if there could be a standard for grading a research paper on readability to raise the status quo.
Also, a lot of work must be shown to demonstrate completeness and rigorousness, and that doesn't help comprehension either.
The trick is to read only the segments of the paper that are actually worth reading, which is often only the abstract, conclusion and maybe implementation details.
Wheels, cars, computer etc.
"Morever, machinery is veiling. It is a way of hiding our inaction from ourselves under what appear to be actions of great effectiveness. We persuade ourselves that, comfortably seated behind the wheels of our autos, shielded from every unpleasant change of weather, and raising or lowering our foot an inch or two, we have actually traveled somewhere.[...] Therefore, the importance of reducing time in travel: by arriving as quickly as possible we need not feel as though we had left at all, that neither space nor time can affect us—as though they belong to us, and not we to them. We do not go somewhere in a car, but arrive somewhere in a car. Automobiles do not make travel possible, but make it possible for us to move locations without traveling. Such movement is but a change of scenes. If effective, the machinery will see to it that we remain untouched by the elements, by other travelers, by those whose towns or lives we are traveling through. We can see without being seen, move without being touched."
the platonic ideal of speedreading: spend 1 minute on each book, read 10k books per year. The shallower the better, complications might negatively impact reading speed.
I can even look away from the screen and still follow along.
It has changed my life and introduced me to so much more information that I would have otherwise not attained with manual reading.
Eye scanning and mental vocalization are really tiring on my brain. Text to speech has solved both of these problems.
I think there’s something about adding a new dimension (auditory) that also helps with memorizing and comprehension. It adds more data points for my Bayesian brain to use and associate with.
Any app/software you can recommend?
I have no trouble believing that speed reading concise, efficient writers represents a loss of information. But some writers just will not get to the fucking point, and those were the books/classes where I was a little jealous of the speed readers. I need to read and retain about 10% of this garbage, but my brain is not cooperating on skimming.
However, if I think about it, reading has been done for thousands of years and by now something like that would have prevailed if it was effective.
Anyway, we need more experiments like this. I bet there are a ton of things out there that we think make out lives better, but in reality don't do a thing.
on an individual basis, there is some evidence that reading with your "best" font may have a meaningful impact on reading speed with no loss of comprehension. but you could probably unlock these gains by simply switching from, say, arial to garamond (i made that up for illustrative purposes), rather than implementing this complex font style.
I can also imagine a version of this which is contextual, based off a query, or a personalized recommendation system.
I spend much more time skimming than reading, and skimming to determine if something is worth reading. Anything that can support that is incredibly valuable and would increase my functional reading speed for accomplishing tasks.
My takeaway is that system fonts, reader mode fonts, etc., should always be configurable.
The basic idea is that you flash words on a screen in the same location so that your eyes don't have to pan back and forth, you just look straight ahead.
Your eyes and brain can recognize whole words as a gestalt (without reading each letter.)
That's it.
This method cuts out most of the physical overhead of reading.
When I played with it I got to the point where I could read over 500 words/minute. I could read faster than my internal voice could speak.
ps- philly native here :)
> Since posting this experiment, I've received a lot of side comments along the lines of, "Well, of course I don't expect Bionic Reading to work for most people, but for [my subpopulation], it really works." If that were the case, we might expect to see disproportionate benefits for those participants who read faster with Bionic Reading than for those who read faster without Bionic Reading. Let's look at how many participants read faster with each font and their average speed gains.
there might be some specialized effect on folks with known reading problems, but bionic reading makes no claims to specifically work for those people but rather the whole population so that's what we decided to test.
Have you been diagnosed with, or do you believe you have, dyslexia, ADHD, visual impairment, or other reading challenges?
It would also be interesting to ask people if they felt it was easier to read with the new tech or the old way. Speed is one metric, but subjective impressions of reading ease are also relevant.
“Average speed gains” doesn’t seem like the best metric here. A graph showing the distribution of users across different WPM differences would be much more informative. Sure, the average user doesn’t see outsized results, but are you saying that zero participants in your test read substantially faster with BR?
That said, I did feel like the bionic rendering made me more likely to read every word as opposed to skimming sentences. It also made reading feel more "percussive" which can be fun. Definitely an interesting line of inquiry even if it's over-hyped.
Does someone know if that exists or if there is research on it?
Maybe I'm just older, have seen more, and have a better tuned BS meter.
SQ4R - Survey/question/read/recite/reflect/review
Gummy bear trick - Place candy/gummy bears on the pages and eat them as you read them.
SparkNotes - Read someone else's summary of what you need to absorb.
First line: tl;dr. Actually no, the results will probably not surprise you
fuck this
I don't know if they still do it, but at the time they were showing off these weird baggies you put over your arms like oversized mittens and ran coolant through.
We know sleeping helps with memory retention. We know that NSAIDs influence experiences of emotional pain. What other folk remedies and completely random things also work on memory? I forget, is caffeine now established wisdom or do they keep going back and forth on that?
I've seen a lot of viral social media posts about bionic reading recently and this is the very first time I've ever seen anyone mention reading speed. Everything I've seen is selling bionic reading for greater reading comprehension and focus. Never mentions speed.
Granted the article measures reading comprehension too (though I'd have some doubts about the methodology of 3 MCQs on a PG article - this part of the test didn't seem high on the author's priority list).
Did the authors just waste a lot of their time because they didn't pay attention to the claims or are we just in very different bubbles?
"Did you know that your brain reads faster than your eye?"
title of bionic-reading.com:
"Faster. Better. More focused. Reading."
does "faster" not refer to speed to you or something?
I guess that answers my question: different bubbles
So in this example: "Bionic reading does not change reading speed: results from our experiment" instead of "Does Bionic reading work? ...".
The change suggested here would be one of the greatest single improvements to the signal-to-noise ratio on this site, and even news in general.
It would be a policy change rather than a tech change really, and would need some enforcement. Just a reminder on the submission page, and perhaps a banner at the top of the home page highlighting the policy change, and directing people to flag titles in the wrong form.
(Or, failing that, does anyone have any suggestions for good tech news sources where the titles are in this format, this is definitely something I want to see)
I'll take my downvotes now.
This is a huge flaw of this study
if you want 'hackers' to read something, best not start with the same line as every ultra-low-quality invasive clickbait ad