Welcome to MSDN Blogs Sign in | Join | Help

ClearType = Clear Mind

In a previous post I talked about the ClearType improvement in word recognition. Lee Gugerty and Rick Tyrrell at Clemson University found that ClearType improves word recognition accuracy by 17%. One of the other studies in the same publication looked at the effect of ClearType on reading sentences. If ClearType improves the accuracy of word recognition then we should see speed and/or comprehension improvements when reading full sentences.

They used a methodology that psychologists have long used to measure sentence comprehension. The participants in the study would read pairs of sentences and determine if the sentences belonged together. Half of the sentence pairs belonged together and half did not make sense together.

Sentences that belong together: 
      Drew picked up the starfish from the sea floor. 
      He was very happy to be scuba diving again.

Sentences that do not belong together: 
      Drew picked up the starfish from the sea floor.
      He was very happy to be sky diving again.

Participants were presented pairs of sentences like the ones above, and pressed one button if the sentences belonged together and another if they didn’t belong together. The computer measured the time from sentences being presented until a button was pressed, as well as the accuracy of the responses. 46 Clemson university students read 144 sentence pairs each.

Overall, the study participants read the ClearType sentences statistically reliably faster than the sentences rendered in black & white (called aliased here). The magnitude of effect is approximately 5%. From the figure it’s clear that while ClearType improves both italic and normal (Roman), ClearType has a larger improvement on the italic text.

Additionally, the participants made more correct responses when the text was presented in ClearType. This is important because it’s very common to see speed improvements at the expense of accuracy, but we’re not seeing that here. There was a statistically reliable accuracy improvement ClearType with a smaller effect magnitude of about 2%.

This sentence comprehension study nicely extends the earlier word recognition finding. ClearType works by making each word easier to recognize, and easier to recognize words results in faster sentence reading and improved comprehension.

Kevin Larson

Gugerty, L., Tyrrell, R. A., Aten, T. R. & Edmonds, K. A. (2004). The effects of subpixel addressing on users’ performance and preferences during reading-related tasks. ACM Transactions on Applied Perception, 1(2), 81-101.

Published Tuesday, December 13, 2005 10:49 AM by fbcontrb

Comment Notification

If you would like to receive an email when updates are made to this post, please register here

Subscribe to this post's comments using RSS

Comments

# re: ClearType = Clear Mind

Can you post any information on the effects of dpi?

My guess is that the difference between a 96dpi screen and a 200dpi (which I am hoping to have in the near VISTA timeframe) at the same size would make an even larger difference, and that clear type would actually be less effective as dpi increases. Still damned useful, and any improvements to display are very valuable.
Tuesday, December 13, 2005 3:28 PM by theCoach

# re: ClearType = Clear Mind

Can you please post an accuracy graph as well? I would like to see how the numbers compare.
Tuesday, December 13, 2005 4:36 PM by Gabe

# re: ClearType = Clear Mind

Hi Coach,
This study was carried out with a 115dpi screen and the operating system set to the default 96dpi. 115dpi is just a little bit higher than average screen density.

I don't have any data for what happens when screen density increases to 200dpi, but can anecdotally say that ClearType is still beneficial at 200dpi. I’m using the 204dpi IBM T221 right now, and while both ClearType and black & white text both look better on this screen, the ClearType text is noticeably better.

It’s true that at some screen density ClearType will become unnecessary, but we are a long way from that time. There are font companies that we work with that make good money hinting fonts for 600dpi printers because hinting is still necessary at that density.

Cheers, Kevin
Tuesday, December 13, 2005 4:54 PM by Kevin Larson

# re: ClearType = Clear Mind

Hi Gabe,

Gugerty et. al. report that the participants answered 91.5% of the ClearType trials and 89.8% of the black & white trials correctly. This is a statistically reliable difference, F(1,45) = 6.06, p = .02. There was no statistically reliable main effect for italics nor a rendering by italics interaction. They did not provide a graph for this data.

Cheers, Kevin
Tuesday, December 13, 2005 5:26 PM by Kevin Larson

# re: ClearType = Clear Mind

What I would like to see is a comparison of ClearType against standard grayscale antialiasing _on a CRT_. I have never liked ClearType on my CRT because it just makes things look fuzzier, and I don't like the red and green fringes. I realize it was designed for subpixel rendering on LCDs and that it works well there, but it seems like you guys are pushing it even for CRT users which seems very strange to me.
Tuesday, December 13, 2005 5:57 PM by Brian

# re: ClearType = Clear Mind

Hi Brian,

I don’t think that Microsoft is “pushing” ClearType on CRTs, in fact a significant number of people feel that we under-promote ClearType on CRTs. As we see it, the performance of ClearType on CRTs depends on your CRT. With Trinitron style aperture grill systems, ClearType works great. On CRTs with misconvergence or poor analog signals it performs poorly (as does most graphics.) For CRTs we’d recommend that you try it out, try tuning with the ClearType tuner and see if it works on your CRT.

Greg
Tuesday, December 13, 2005 7:49 PM by fbcontrb

# re: ClearType = Clear Mind

Sorry for the repeat, but I previously posted this question in an old old post that may not be read anymore. I love the blog, and I love ClearType. But at work I use a portrait display. Now that I am addicted to ClearType at home, I want to get it working in portrait too. Does anyone have any information about how to try to get this to work??
Wednesday, December 14, 2005 11:36 AM by James Kreines

# re: ClearType = Clear Mind

James: One of the ClearType design specs is that it works with certain RGB orderderings in the pixel components. This doesn't work very well when you rotate your pixels 90deg. And since your video card driver is doing the rotation, you're already a level removed from the ClearType APIs. (I may be wrong on this, but that's how I've understood ClearType works.)

FontBlog: ClearType might improve word recognition, but 2 and 5% are hardly much of an endorsement. What I'd like to see is a study of ClearType fans to determine whether they can tell the difference between CT and AA text on a variety of monitors.

(I like ClearType, but I'm not an evengelist like some CT fanbois I know. CT has some disadvantages; namely, it's slower, it's not as portable, and selected (inverted) CT text looks like a rainbow font.)

p.s. Is there something wrong with the grayscale hinting on the em in this blog's font. (Calibri?) There's an excessive trailing space after the em that really breaks up words and IMO looks strange.
Thursday, December 15, 2005 11:27 AM by Dan McCarty

# re: ClearType = Clear Mind

James: I think Greg is going to answer your portrait mode question. The full answer is fascinating and surprising.

Dan: I strongly disagree, a 5% speed improvement with a simultaneous 2% comprehension improvement is an important finding. If on-screen reading was a task that only a few people did, a couple times a year, and only took a few minutes, these improvements would be meaningless. But reading is arguably the most frequent task you perform at a computer. Nearly all tasks that people perform with computers involve a heavy reading component, including tasks that people may not realize involve reading such as studying a spreadsheet and writing a document or email.

ClearType improves the speed to complete all the reading involved tasks that you perform on screen. While it obviously doesn’t double your reading speed, a 5% speed advantage saves you 3 minutes for every hour you spend reading. If you spend a lot of time at a computer, this can quickly turn into a big time savings. Jakob Nielsen once tried to monetize this benefit.
http://www.useit.com/alertbox/20020203.html

Cheers, Kevin
Thursday, December 15, 2005 3:09 PM by Kevin Larson

# ClearType in Internet Explorer 7

Hi. My name is Bill Hill, and I’m one of the original inventors of ClearType. My job at Microsoft is...
Friday, March 03, 2006 4:02 PM by IEBlog

# re: ClearType = Clear Mind

Can the researchers test the same font in print vs. screen to get a notion of the reader's basic transitional capabilities?

Can the researchers test Roman Vs. Italic, (in the main event!) so we can get an idea of the coprehensional differences between them?

Can we get all of the type solutions on a user's machine bundled into single tuner, that would prove you loved us and were not pushing anything but choice. ;)


Monday, March 20, 2006 11:25 AM by David Berlow

# re: ClearType = Clear Mind

Hi David,

John Gould at IBM ran a series of studies comparing print to screen. Display quality has improved some since he ran his studies, but I suspect that the findings are still useful.

This study is an example of comparing Roman to Italic, and matches the findings of many previous studies that have looked at the Roman v. Italic question. Tinker & Patterson's work is well known, and there are others. Though this study only looks at Verdana and Verdana Italic, and I'm not sure I'm ready to generalize these findings to all italic fonts.

I'm not sure what you mean by a single tuner for all type solutions, but of course we love you David. Could you elaborate?

Cheers, Kevin
Monday, March 27, 2006 6:17 PM by fbcontrb

# re: ClearType = Clear Mind

"what you mean by a single tuner for all type solutions,"

Hi Kevin! I mean, a tuner, that allows the user to select from all the resident rasterizers, and fonts, and sizes, to get to the perrrrrrfect reading for them.

I mean, I think everybody agrees that no test shows 100% of anybody wanting any one solution to the screen font "thing." Do you know of a test where even 80% prefer CT over another reasonable solution?
Monday, April 17, 2006 11:41 AM by David Berlow

# re: ClearType = Clear Mind

Great
Saturday, April 29, 2006 4:20 PM by tecsi

# re: ClearType = Clear Mind

I'd like to see a fatigue test - for example, reading faster is no use if you can't do it for as long as you would normally.

I say this because I find ClearType very tiring to look at (and aesthetically unpleasant) because of the inconsistancy in contrast and variance in colour.

Vista does look better 'out-of-the-box' than XP on my monitor, but is still somewhat soft and faded.  

That combined with the fact that Vista often requires you to read full paragraphs of information for each dialog, rather than spot short labels at a glance, adds up to a serious usability problem for me.
Monday, August 28, 2006 6:46 AM by Paul Coddington

# You say it 'looks good on paper?' It must not be using ClearType....


I have posted before on this issue, several times, in fact. And so have many others.
But now there...
Monday, September 11, 2006 3:34 AM by Sorting It All Out

# re: ClearType = Clear Mind

It is very surprising that 2% can be statistcally reliable. Anyway, I have a Phillips 17" monitor, and cleartype at IE looks really bad (everything looks fuzzy).

Saturday, October 07, 2006 1:14 PM by Pablo

# re: ClearType = Clear Mind

I hope the MS updates ClearType (to version 2.0?) to allow Portrait rasterizing.

Currently, ClearTpye increases the "x" resolution by utilising the sequence of RGB. The "y" resolution remains the same.

It is equally straightforward to have a portrait mode.  However, due to the alignment of the RGB on portrait mode, the "y" resolution would now increase while the "x" would remain the same.

For scan-line reading (most languages) the rotated Cleartype may not be quite as effective... however it will still be better than one.

For scan-column reading (chinese?) the rotated Cleartype may be better then standard cleartype.

However, what we really need is a higher resolution TFT to start with, operating at 150dpi or better. This would help significantly.

Tuesday, October 10, 2006 9:48 AM by David

# re: ClearType = Clear Mind

Before today I knew nothing of ClearType.  I have three 19" LCD's.  I turned on ClearType an hour ago and feel that 5% is an underestimation.  Mental anguish while reading my screen has also considerably decreased.

Wednesday, December 13, 2006 1:23 PM by Brian O

# Dubroy.com/blog » Sub-pixel font rendering: it ain’t just about the looks

# Font Rendering: Respecting The Pixel Grid

I've finally determined What's Wrong With Apple's Font Rendering . As it turns out, there

Friday, June 15, 2007 12:54 AM by Programming

# » Blurry, but fast: Safari’s fonts | Between the Lines | ZDNet.com

# Not all developers like fixed width fonts

In particular, I don't. Lots of times other developers come over to me and look confused as they see

Tuesday, July 03, 2007 8:07 AM by CodeObsessed

# re: ClearType = Clear Mind

I'd like to see this same test run on Apple users. I am concerned with the apparent disregard for the fact that the people chosen may have just been more -familiar- with ClearType rendering than with fuzzier rendering.

Wednesday, July 11, 2007 2:05 AM by allo

# Stoneburial › Digital font display

Monday, March 31, 2008 8:57 PM by Stoneburial › Digital font display

# ClearType improves the efficiency of typical office tasks

In two earlier posts I talked about studies that showed that word recognition is 17% more accurate and

Wednesday, July 16, 2008 3:10 PM by fontblog

Leave a Comment

(required) 
required 
(required) 
 
Page view tracker