Extensive reading and reading speed

[phills]

With the recent discussion on reading speed, I was inspired to run some tests.  I decided to compare my English v Chinese reading speed. 

 

As content, I chose Chapter 3 of The Magician's Nephew, a short easy piece I'm already familiar with, that I have available in .txt form in both Chinese & English.  I read it years ago as a child in English, and I read it in Chinese less than 6 months ago.

 

In English, Chapter 3 has 2809 words.  In Chinese, Chapter 3 has 4627 characters (so roughly 1.65 Chinese chars per English word)

 

I read the English version in 6.82 minutes, for an average of 420.97 words per minute. 

 

I purposefully read in a leisurely manner, trying not to scan / skip words, but not reading word-by-word like a child either.  I think that's faster-than-average English reading speed, but not close to a speed-reading level of speed.

 

---------

 

Then I read the Chinese version right after the English version.  (Note: I've read this exact text less than 6 months ago, and I just read the same text in English 10 minutes ago, so the meaning of every paragraph is fresh in my head.  It's about as perfect a situation as I can get in Chinese.)

 

I read the Chinese version in 16.56 minutes for an average of 279.41 characters per minute.  That's by far my speed record.  My normal speed is around 150 chars per minute, +/- 25. 

 

Because I knew what every sentence was supposed to mean, I spent almost no time on processing logic.  Just recognizing characters, parsing the sentences into words and deciding that the words match what I knew they were trying to say, so I can move on. 

 

Even than my average English speed is 2.4x faster (6.82 mins v 16.56 mins) than my max Chinese speed.

 

My average Chinese speed is ~150 cpm (my max Chinese speed is 86% faster than my average speed).  And that it makes my average English speed 4.5x faster than my average Chinese speed.

 

I knew I had a ways to go, but now I know exactly how wide the gap is   4 to 5x. 

 

I had been thinking of pulling double-duty by choosing to read Chinese versions of  novels / texts I've put off reading in English.  I'll still do that, but I can see it won't be saving much time (but still good for Chinese language development)! 

 

Anyways, even though the test is specific to me & 1 piece of text, I thought the data (particularly the ratios) might be interesting to others.

[pinion]

On 10/2/2021 at 12:57 AM, 大块头 said:

It could be, but I think total reading time is a better choice of independent variable. The model I've come up with assumes that the rate at which your reading rate increases is linearly proportional to the difference between your current reading rate and the native reading rate your performance will asymptotically approach.

I'm curious what makes you feel that total reading time is the better independent variable. My instinct is that it really is characters read that is the determining factor--that is, for a given unit of time spent reading, you get more benefit from reading something easy and fast (i.e. more characters read) than something hard and slow.

[pinion]

@phills Glad my stats have been useful for you! Unfortunately all my data since I made this post has been affected by a couple of changes (switching from simplified to traditional, switching to reading primarily on paper), so there probably won't be much more in the way of charts from me...so I'm delighted that other people are adding their data to this thread.

 

Re: English vs Chinese reading speed: I don't have as convenient a sample for comparison as you do (I try to avoid reading translated work), but your stats ring very true to me. Getting my Chinese reading speed to catch up to my English reading speed is probably an impossible dream, but I'd like to at least get to...0.5x, maybe? That seems reasonable.

[phills]

	@pinion if you can think of a good way to compile it, I'd like to see the effect on speed of switching to traditional.  I tried it 1 million chars ago, and it disrupted my reading speed quite a bit, causing me to second guess things I used to do without problems.  Of course, I didn't think of measuring it at the time.    Next time I try it, at 5 million chars, I'll see if I can capture the effect in data.   As for switching to paper... I can see how that'll end the easy data collection effort
	On 10/5/2021 at 1:13 PM, pinion said: Getting my Chinese reading speed to catch up to my English reading speed is probably an impossible dream, but I'd like to at least get to...0.5x, maybe? That seems reasonable.   Ya, half English speed would be nice long, long term goal.

[pinion]

On 10/5/2021 at 2:24 PM, phills said:

if you can think of a good way to compile it, I'd like to see the effect on speed of switching to traditional.

That I can actually answer! Here's the original graph with the addition of my early adventures in traditional characters:

 

 

The non-blue dots on the right represent books I read in traditional. The three red dots are all part of the 盜墓筆記 series; the three pink dots are the three books of 三體. Note that the very first book I read in traditional characters is not included--I deliberately didn't track time for it in an attempt to avoid putting too much pressure on myself. All told, by the time I got to the green dot I'd read about 1.5 million characters in traditional and gotten back up to a speed of about ~225 cpm, but the initial drop when switching simplified>traditional was very dramatic indeed.

[phills]

Thanks @pinion.  Very interesting as always. 

 

I expected a very large drop in reading traditional characters when you first start, but did you also read simplified characters at the same time?  Meaning switching between the 2?  ( I was reading a "Condor Heroes" comic book in traditional, while at the same time reading the novel "三体 2"in simplified).

 

I found even my reading simplified characters slowed significantly, because I kept confusing characters in my head in a way I didn't before I learned traditional.  When I dropped the comic book, and returned back to 三体 2 full time, my speed restored itself by the end of the novel.

 

Maybe the way is to switch to traditional for a while, and not switch back, until you've stabilized your knowledge of traditional characters in your head.

[phills]

I often observed that reading gets easier as you get further into a book, so I wanted to quantify it.  To see how reading speed varies throughout a novel, I measured my speed reading through Red Finger 红手指. 

 

红手指 is a 2010's Japanese detective novel, translated into Chinese.  It has about 100k characters.  I roughly divided it into 5 parts:

 

1st 20%: 169 cpm

2nd 20%: 175 cpm

3rd 20%: 192 cpm

4th 20%: 211 cpm

last 20%: 221 cpm

-------

Average: 193 cpm

 

Few notes:

 

1. My last novel before this, 黄金时代, I read at 149 cpm. 

 

I read 红手指 30% faster, so by the speed metric it's roughly 30% easier.  Not surprisingly, the type of novel really impacts the speed.  红手指 is a mass market novel, vs the more literary 黄金时代.

 

Could it be that I'm just a much better reader than when I read the last novel?  Nope. 

 

As a baseline, I've also been reading short little chapters of a children's history book on the side.  Before I started 红手指, I read Chapter 25 at 165cpm.  After I finished 红手指, I read Chapter 26 at 164cpm  

 

Overall speed just doesn't budge that fast.  It's the difficulty level of the material that's the more sensitive variable.

 

2. In the first 20%, I was slowed down because of (a) unfamiliar names, particular Japanese names (how do I recognize it's a name, where do they break, how long are they), and (b) unfamiliar author and genre vocab (it's the first time I read a detective story in Chinese). 

 

For a sense of the effect, I collected 54 vocab words from the book, about 30 (more than half) of which came from the 1st third.

 

Also, because it's a detective story, the first bits had a lot more scene setup, while the rest has a lot more dialog (interviews with witnesses, planning/coverup conversations).  Dialog goes fast, as you see in the speed for parts 4 and 5.

 

3. So for a fairly favorable story, you can be reading the last bits at up to 30% faster than when you started.  That puts some numbers to the feeling of speeding-up that I get, when reading through some books.

[imron]

On 10/8/2021 at 7:36 PM, phills said:

Maybe the way is to switch to traditional for a while, and not switch back, until you've stabilized your knowledge of traditional characters in your head.

This would be my advice. Or at least don’t be reading another book in simplified at the same time. 

[Woodford]

I think I'm currently at around 100-125 characters per minute (it takes me 5 minutes to finish a page). There are all sorts of quirks in my reading speed. I read much more slowly at the beginning of my reading session, and then I get much faster once I've read a few pages and "warmed up." I zone out, I daydream, and if the story is good, it sometimes causes me to sit back and think. I often go into autopilot and realize I haven't really processed the previous sentence or paragraph, so I go back and re-read it. I have the bad habit of sub-vocalizing the words, ever since the days of graded readers 3 years ago, when I read out loud to practice my tones/pronunciation. I pretty much never stopped that.

 

I think the somebody earlier on this thread said that he subvocalized until he realized that his reading speed began to outpace his vocalization speed. That's a great milestone! I'm just a slow and relaxed person in general, so I guess my goal is to "be able to read quickly...when I want to!"

[phills]

On 10/9/2021 at 10:59 PM, Woodford said:

I think I'm currently at around 100-125 characters per minute (it takes me 5 minutes to finish a page). There are all sorts of quirks in my reading speed. I read much more slowly at the beginning of my reading session, and then I get much faster once I've read a few pages and "warmed up."

 

I need to warm up too.  The first thing I read in Chinese every day, I feel like I stumble over 25% of the characters, not even recognizing them.  I'm best at night, just before I go to sleep, which is when I do most of my reading.

 

But you should measure yourself.  You might be faster than you think.  I used to think I was consistently below 150, except on good flow, until I measured myself.  Now I think I'm consistently above 150, except on tough material.

 

Of course, one loophole is computer tools count punctuation / spacing.  So I might actually be below 150 in terms of characters but in terms of unicode-chars, I'm above   A clever person might note that dialog has a lot of punctuation / spacing.

 

Another trick I picked up is to only vocalize every 3rd word, or only characters at the beginning / end of a sentence.  I think that's a halfway accomodation to not vocalizing at all, which I can't do yet, so I only choose to pronounce a few of them.  That seemed to help.

 

Finally, I read in an editor.  I notice I tend to space out on long paragraphs.  So I manually insert paragraph breaks every few sentences.  That may be another form of cheating, but it keeps me from having to re-read. 

 

I try to insert the breaks where there is a logic break, so maybe it's a way to engage my brain to keep track of the topics of the sentences.  I also insert "* * *" division breaks, every change of scene, so it's easier for me when I flip back to old stuff.

[Guest realmayo]

On 10/9/2021 at 5:20 PM, phills said:

computer tools count punctuation / spacing

 

A characters-only count is typically 85% to 90% smaller than a count of characters + punctuation, I reckon.

[Woodford]

On 10/9/2021 at 11:20 AM, phills said:

But you should measure yourself.  You might be faster than you think. 

 

Well, I just read a Chinese New York Times article, and I used find/replace in a word processor to take out all the spaces and punctuation. My speed turned out to be 129 characters per minute. So when I'm not trying to read efficiently and I'm just casually reading a book, 75-100 characters a minute is probably my usual speed. I'm sure that baseline value will increase as I pound through more books this year. It might be interesting to try again later.

[pinion]

On 10/9/2021 at 10:59 PM, Woodford said:

I have the bad habit of sub-vocalizing the words, ever since the days of graded readers 3 years ago, when I read out loud to practice my tones/pronunciation. I pretty much never stopped that.

 

I think the somebody earlier on this thread said that he subvocalized until he realized that his reading speed began to outpace his vocalization speed.

I think this is probably a reference to when I said the following:

On 10/12/2020 at 5:52 AM, pinion said:

I read roughly the first 2.5 million characters either fully out loud or muttered under my breath, and switched to reading silently only when reading out loud began to noticeably slow me down.

I want to be very clear, though, because I have very strong feelings about this that run contrary to a fair bit of advice I've seen--I think subvocalization is GOOD and should NOT be actively suppressed. At a certain point my reading speed outpaced my vocalization speed because frankly my mouth isn't very agile and I don't talk that fast, but I am very conscious of deliberately continuing to subvocalize in my head. I think it's crucial to maintain a very strong character-sound connection and to avoid falling into the non-native trap of thinking you can derive meaning from characters without first passing through sound, which is after all the most fundamental part of language.

 

The degree to which you subvocalize (or are conscious of doing so) will shrink as you become more comfortable with reading, but it never completely goes away, nor should it--you just become able to subvocalize at faster than speaking speed, in much the same way as a podcast is still comprehensible at 2x or even higher speed once you're used to it.

[pinion]

On 10/8/2021 at 4:36 PM, phills said:

I expected a very large drop in reading traditional characters when you first start, but did you also read simplified characters at the same time?

 

No, I did not. (I was in a hurry to get up to speed with traditional characters as quickly as possible.)

 

On 10/8/2021 at 4:36 PM, phills said:

Maybe the way is to switch to traditional for a while, and not switch back, until you've stabilized your knowledge of traditional characters in your head.

 

Yes, I think that's probably best. My (dubiously supported) theory of learning is that above and beyond the time you spend studying, you want to convince the subconscious part of your brain that whatever you're working on is important, so that it keeps churning away at the learning process even while you're doing other things, sleeping, etc. And the way you convince your brain that something is important is by exposing your brain to it as much as possible, especially at the beginning, in order to show it that the new thing is here to stay and there's no retreating to or substituting previous knowledge.

 

On 10/8/2021 at 8:21 PM, phills said:

So for a fairly favorable story, you can be reading the last bits at up to 30% faster than when you started.  That puts some numbers to the feeling of speeding-up that I get, when reading through some books.

 

This rings true for me too! I occasionally did reading time projections as I worked my way through a book, i.e. if I'd spent 4 hours reading the first 20% of a book, extrapolated forward one would expect the whole book to take 20 hours. But my projections tended to drop as I got further into the book, so by the time I got through 50% I might expect the same book to only take 18 hours total. Now I'm regretting a little that I didn't keep that data as well.

[phills]

On 10/10/2021 at 10:15 AM, Woodford said:

Well, I just read a Chinese New York Times article, and I used find/replace in a word processor to take out all the spaces and punctuation. My speed turned out to be 129 characters per minute. So when I'm not trying to read efficiently and I'm just casually reading a book, 75-100 characters a minute is probably my usual speed.

 

I find Chinese NYT very hard to read.  Unless you're reading a lot of it, I think you're probably reading novels faster than that, particularly mass market ones. 

 

From my tests, you don't read slower when you're enjoying reading, if anything you tend to read faster (although not that significantly, difficulty is a far bigger factor).  Finally, punctuation adds another 10-15% to your speed, so 129 -> 140-ish. 

 

The reason I'm second-guessing your numbers is because I think @pinion's graph is surprisingly representative, and every body who reads X characters is relatively in the same band of speed.  I'd love to get more people's data for confirmation / dis-confirmation.

 

I think even native speaking children read > 5-10m characters in K->12, so even they are (on an order of magnitude) learning at the same rate.  I'm been googling to try to find confirmation for that, but it's hard to find data on how much kids actually read on average, in any language.

[imron]

I once measured the reading speed of a friend’s 16 year old son. His speed was around 750 cpm, on fictional writing (the Chinese translation of lord of the rings). 
 

There’s huge variance in reading speed though among native speakers, anywhere from 300-950cpm among people I have measured. 

 

 Normal (i.e. not news presenter) speaking speed is 200-250cpm. 

[phills]

At the top end, there's clearly a big variance in speed.  But there's just as huge variance in amount of characters read. 

 

Voracious readers might read 100 million chars vs someone who went through minimal schooling might have read 1 million.  I don't know if those are the precise numbers, but it would be interesting to see them, if they exist.

 

Maybe I'm wrong and you can zoom up the speed ladder without actually having read 10's of millions of chars, but right now my hypothesis is the opposite. 

 

Kids would be the best way to determine this, because I tend to believe kids can learn to go fast, better than adults.  Adults can sometimes learn as fast as kids, but they can't learn to do things fast, as well as kids learn to do things fast.

 

Right now as an adult, I'll never get to 750cpm in Chinese, but I did get to something similar in English as a kid.  But I read a lot as a kid, and I bet your friend's kid has read a lot too.

[phills]

I've been googling on average reading speeds, and the best article I've found is this.  It's a meta study of reading speeds, reviewing 190 other studies. 

https://www.sciencedirect.com/science/article/abs/pii/S0749596X19300786

 

(non-paywall version) https://www.researchgate.net/profile/Marc-Brysbaert/publication/332380784_How_many_words_do_we_read_per_minute_A_review_and_meta-analysis_of_reading_rate/links/5cb0b402a6fdcc1d498feb2d/How-many-words-do-we-read-per-minute-A-review-and-meta-analysis-of-reading-rate.pdf

 

It's got a section comparing the speed of reading in English vs other languages, and it's also got a section about L2 reading speeds (non-native reading speeds). 

 

I thought I'd save down some tidbits for my own benchmarking purposes:

 

1. "Based on the analysis of 190 studies (17,887 participants), we estimate that the average silent reading rate for adults in English is 238 word per minute (wpm) for non-fiction and 260 wpm for fiction."

 

2. "Reading rates in other languages can be predicted reasonably well be taking into account the number of words these languages require to convey the same message as in English."

 

3. (p. 49)  "In Chinese we found 26 studies coming from 23 articles: 18 on silent reading and 8 on reading aloud.  Mean reading rates were respectively 260 wpm and 152 wpm. With respect to these estimates, it is important to know that we often had to estimate the number of words from the number of characters given. When we had to do so, we used the conversion 1.5 characters for 1 word. For those texts in which the words on average were longer than 1.5 characters, our value is an overestimate... "

 

Native speed benchmark: 260 wpm = 390 chars per minute, at the exchange rate of 1.5 chars = 1 word.

 

4. (p. 31)  "Reading speed in second-language (L2) speakers is considerably slower than in first-language (L1) speakers. Indeed, reading rates below 100 wpm are no exception. Hirai (1999), for instance, studied English L2 reading rate in Japanese university students. All students had taken six years of formal English education in junior and senior high school. In addition, most of the participants had two 90-minute English lessons per week at the university and a subgroup majoring in English had about five to eight English courses per week. Text materials were easy prose passages, followed by a set of eight four-option multiple-choice questions. Reading rate was 139 wpm for participants who could answer more than 75% of the questions correctly and 61 wpm for the other participants. Interestingly, Hirai (1999) also tested the participants on English L2 listening and found that their estimated optimal listening rates corresponded well to the observed reading rates. "

 

L2 learner benchmark: For L2 university students who learned the second language fairly well (including language majors), the equivalent average would be 208 cpm, for Chinese characters.  For those who didn't (but still sat thru the mandatory 6 years of secondary school classes), the equivalent average would be 91 cpm, for Chinese. 

 

L2 learners typically listen at approximately those rates as well.

 

5. (p. 31) "Cop, Drieghe, and Duyck (2015) asked reasonably proficient Dutch-English bilinguals to read half a novel in L1 and the other half in L2. Reading rate was 17% slower in L2 than in L1. In addition, the eye movement pattern of L2 readers very much resembled that of L1 children: They made more fixations per sentence, fixations times were longer, forward saccades were shorter, and less words were skipped. Only the number of regressions did not differ. Similar results were published by Whitford and Titone (2014) for sentence reading, although in their study regression rates were higher in L2 than in L1 as well. Dirix et al. (2019) observed 10% slower processing rates when participants read or studied texts in a second language (respectively 174 wpm and 50 wpm) than in the first language (189 and 54 wpm). "

 

Proficient bilingual benchmark:  17% slower than an average of 260 wpm is 216 wpm.  That would be an average for proficient bilinguals of 312 chars per minute, for Chinese characters.

 

[Guest realmayo]

On 10/8/2021 at 1:21 PM, phills said:

I often observed that reading gets easier as you get further into a book, so I wanted to quantify it

 

To add a datapoint, here's a record of the five days I spent reading 余华's 许三观卖血记. For what it's worth, the character count excludes any punctuation etc, and the time includes looking up words etc.

 

hh:mm	字	speed/cpm
1:31	10,924	120
3:19	27,231	137
2:40	22,352	140
1:55	17,532	152
2:19	20,170	145

 

Farily similar % difference from slowest to fastest compared to phills. My last chunk was not the fastest -- I find that towards the end of a book I either accelerate, because I'm bored and want to get it finished, or slow down a bit, to enjoy it & make sure I take everything in.

 

Oh, and this was the cpm speed for 兄弟（上）, which I also kept track of a while ago:

 

110

134

153

137

162

141

 

[alantin]

On 10/13/2021 at 4:56 PM, phills said:

5. (p. 31) "Cop, Drieghe, and Duyck (2015) asked reasonably proficient Dutch-English bilinguals to read half a novel in L1 and the other half in L2. Reading rate was 17% slower in L2 than in L1. In addition, the eye movement pattern of L2 readers very much resembled that of L1 children: They made more fixations per sentence, fixations times were longer, forward saccades were shorter, and less words were skipped. Only the number of regressions did not differ. Similar results were published by Whitford and Titone (2014) for sentence reading, although in their study regression rates were higher in L2 than in L1 as well. Dirix et al. (2019) observed 10% slower processing rates when participants read or studied texts in a second language (respectively 174 wpm and 50 wpm) than in the first language (189 and 54 wpm). "

 

Interesting. Maybe I'll try measure my reading speeds sometime. This sounds about right to me. I expect my English reading to be somewhat slower than my native language, though I do most of my reading in English.