DIY reading speed test
As regular Typophile readers are aware, there has been much discussion over the past many months regarding reading, readability, and the testing of reading speed and accuracy. One of the questions that has come up again and again is what constitutes typical reading speed for mature, experienced readers in what might be called natural conditions, i.e. not laboratory conditions. I would like to invite your participation in an informal, self-regulated study. Obviously the fact that it is self-regulated means that the results are of limited scientific value, but I think as an initial step toward finding a good means to test what Hrant called 'immersive reading' the process may be a useful one.
In order to participate in this test you will need a) a book, b) a watch, and c) an assistant. The book should be something you are reading anyway, and preferably you should be some ways into it before conducting the test. It should not be a book that you have already read: the text should be new to you. A stopwatch can be used, but not if the click of the stopwatch starting distracts you during reading. The assistant holds the watch and is located someplace where he or she can see you clearly but not where his or her presence will distract you. You should read in the situation in which you most like to read: in your comfy chair, for instance. You should try to conduct the test when you are well rested and your eyes are not fatigued.
This is how to conduct the test:
1. Start reading. The assistant will watch you and will not do anything for the first while. The assistant will need to judge when you are immersed in the text and, judging by how regularly you turn the pages, that you are reading at a reasonably constant speed.
2. Selecting his or her moment, as you turn a page, the assistant starts timing you. If not using a stopwatch, the assistant should write down the starting time. The assistant should also keep track of how many times you turn the page during the timed period, so that you can accurately determine where in the text timing started.
3. The assistant times you for ten minutes. At the end of ten minutes, the assistant calls 'Stop', at which point you put your finger in the text to mark the part you have reached.
4. Calculate the number of words you have read during the ten minute test. It is not essential to have an absolute count: calculate the average number of words per line on a single page, and then multiply that by the number of lines per page and the number of pages read. Count only the number of completely read lines on the last page, down to the point where the assistant called 'Stop'.
5. Post the results here. Results should include:
a) Type of book read (novel, biography, history, etc.);
b) Your sex -- if your gender is different from your sex, I don't want to know about it 
;
c) Your age range (20-30, 30-40, 40-50, etc.) or actual age;
d) Number of words read during the ten minute test.
I'll be carrying out the test myself when my influenza clears.
20.Dec.2004 1.02am
IN any langues? any scripts?
20.Dec.2004 7.39am
This is a good idea, but it's totally important than the reader forget he's being tested*. This requires for example than the assistant not be visible. Also, ten minutes does seem long enough, but it assumes that "the assistant will need to judge when you are immersed in the text" is a reasonable expectation. Lastly, it's probably better not to put an upper limit on time, and let the assistant decide arbitrarily, otherwise the expectation of being stopped would limit immersion; or you could let a "random" event (like a telephone call) determine the stop point.
* This being the essential flaw in virtually all empirical testing of readability.
BTW, it might be useful to use a camcorder (placed behind the reader), and take some time after the session to decide where to start measuring.
--
You're going to need a lot of data to get any useful insights, partly because letting the reader choose the book throws a wrench into the "readability quality" of the text. If we're measuring at what speed people typically read, such variance is a good thing; but if we're trying to figure out how people read as immersively as possible -in order to make fonts better- then we need to control the reading material.
Oh, and where's Kevin, dude?
hhp
20.Dec.2004 2.44pm
Good points, Hrant. I've suggested that the assistant be located where he or she will not distract the reader. Obviously this means at a distance and not in the reader's peripheral vision. The video camera idea is a good one, but I wanted to keep the test as simple as possible, so that lots of people can participate. My interest at this point is typical reading speed, so I decided not to try to control the textual variables. People typically read things that they want to read, so providing them with a set text introduces the kind of artificiality that you find questionable in much empirical testing. The idea of encouraging people to read from a book that they have already started is that this is, presumably, something that they want to read.
By the way, people can perform the test more than once, posting results from multiple sessions, either with the same book or with different books. I'll collate such data slightly differently from individual submissions.
Oh, and where's Kevin, dude?
I've notified all our colleagues from the legibility panel in Thessaloniki: Kevin Larson, Peter Enneson, Mary Dyson, and also Ole Lund. I've also invited ATypI members to take part, and Jean-Fran
20.Dec.2004 3.02pm
> providing them with a set text introduces the kind of artificiality
> that you find questionable in much empirical testing.
Not set text (since I totally agree that using something the reader wants to read is indeed a great facilitator of immersion), but set typography. So ideally the reader would pick the book and you'd give him a good setting of it. But sure, that's not practical [here]. Maybe one day when average screens are good enough that can be done though.
Anyway, I hope it takes off!
hhp
20.Dec.2004 9.49pm
>I'm hoping Simon Daniels will publicise it on the Microsoft Typography site.
sorted...
http://www.microsoft.com/typography/links/news.aspx?NID=4471
Cheers, Si
20.Dec.2004 10.04pm
Ok, here are my results. I read 10,774 characters of Ernest Hemingway
21.Dec.2004 12.26am
H
21.Dec.2004 12.31am
Jean-Fran
21.Dec.2004 2.04am
H
21.Dec.2004 4.20am
So, all people reading English not as their first language need to specify it also for the results.
21.Dec.2004 4.25am
Yes, although again, I probably won't collate that data if it looks likely to throw out the overall results. What I'm interested in is typical reading speeds for 'mature, experienced readers', which means experienced readers in the test language. I'd be more interested to know how fast you read French, than how fast you read English.
21.Dec.2004 9.39am
Comparing reading in different languages and scripts would surely be extremely revealing*. Measuring reading of Chinese versus English for example would give us huge clues - but it would also be extremely difficult to read those clues properly. In my mind the best such comparison -considering the current immaturity in readability testing- would be between English and German, since the main difference between them is word length. I remember proposing this in Thessaloniki; it's certainly something MS could pull off - but not if they insist on using ClearType for the testing...
* Not least because such differences actually seem to cause differences in brain structure development! (As per that recent BBC-Radio series on reading.)
> 'mature, experienced readers'
Yes, which means you also should collect information about the subjects' reading experience. Since this is hard to quantify, you could substitute education level, plus something like the number of books read per year, and age. Hmmm, does Typophile count? :->
The thing is, John, what exactly do we expect to learn from this? I think Kevin would be the first to point out that this is really way to informal to draw practical conclusions. Which fortunately doesn't make it pointless, but one thing it does mean is that the amount of data you collect has to be exponentially greater than in a formal study. Sort of how anecdotal evidence is "collected".
BTW, the average English word length is closer to 4. For Spanish it's closer to 5.
hhp
21.Dec.2004 11.19am
John, your reading speed test is a fabulous idea. It will be very interesting to learn people
21.Dec.2004 11.35am
> If we find the 700 words per minute that Hrant has hypothesized
But you'll only find it* when the conditions are right. You might think what's the point of using such a narrow -and quite atypical- reading environment, but again the intent (to me at least) is to see how the machine works, not how fast it's typically driven. When we figure out the innards then we can improve our fonts. Knowing typical behavior doesn't help much with pushing boundaries.
* BTW, 700 is just something I threw out, an educated (maybe) guess.
I think at the low end (deliberation, or light immersion) you would expect the speeds to be comparable, and I think the few test that have been done actually show this. But as you know my contention is that the top "half" is different, and different scripts would start diverging a lot. Hebrew versus* Arabic would be another interesting pair, since they're both right-to-left and consonantal, but one is very enclosed while the other is very open**. My guess is Arabic is more conducive to immersion, while Hebrew has better apparent size.
* Sans bloodshed though!
** Envelopes being critical in parafoveal reading.
hhp
21.Dec.2004 1.59pm
John, I got the count of 10,774 characters on a printed book by counting the average of characters on a full length line and then multiplying this average by the amount of full lenght lines, the same I did for the 3/4 lenght lines, the 2/4 lines and the 1/4 lines. Then I added the results and got that number.
I got an average of 13.2 words per full length line and an average of 69.5 characters by line so we have an average of 5.26 characters by word (without substracting the spaces).