AGENCE FRANCE PRESSE v. GOOGLE INC.
RULE 26b4 STATEMENT. (Attachments: # 1 Supplement Part 2# 2 Supplement Part 3)(Elgin, Jennifer)
AGENCE FRANCE PRESSE v. GOOGLE INC.
Doc. 60 Att. 1
Many of us can read the brass insignia and quickly understand the first layer of information: Sergeant.
Page 1 of 8
Page 38: To many Americans this represents the "Gold Star Mothers" organization and indicates that the wearer has a family member serving in a combat zone. It's impossible to substantiate this claim without a quantitative study. I didn't recognize the Gold Star Mothers organization in the small lapel pin and I doubt that most Americans would.
We have to use our visual acuity to read the content of the insignia and understand that this is a Sergeant First Class - E7.
Going beyond this level of information we see a red bordered rectangle with a blue star on a white field. To many Americans this represents the "Gold Star Mothers" organization and indicates that the wearer has a family member serving in a combat zone.
This small lapel pin we recognize as containing the National colors.
Page 39: To gain more and pertinent information, we look at the
At the first layer of information we recognize this as a postage stamp. To gain more and pertinent information, we look at the monetary amount the stamp represents, 37 cents.
monetary amount the stamp represents, 37 cents. The next layer, the least important to many users, is what the image represents, in this case, Lewis and Clark. The qualitative criteria that Mr. Nesbitt misses in this case, and in many of his examples, is context. In evaluating this example, Mr. Nesbitt places higher value on the denomination of the stamp than the image it commemorates. But a numismatist or historian may place higher value on Lewis and Clark the very information Mr. Nesbitt dismisses as least important. We cannot assume that all readers and users process information using the same hierarchy of importance. Pages 40-41: The next example is a magazine page of magazine pages.
The next layer, the least important to many users, is what the image represents, in this case, Lewis and Clark.
Having noticed that the stamp is worth 37 cents and today's postage requires 39 cents, I have to rummage through the desk drawer to find an even smaller stamp and read its value properly, 2 cents.
The next example is a magazine page of magazine pages. Each subordinate page is filled with content information that may or may not be informative, helpful or useful. To discover that, we have to read the content of the individual pages on the page. As small as they are, it is still quite easy for us to read this content and comprehend.
Each subordinate page is filled with content information that may or may not be informative, helpful or useful. To discover that, we have to read the content of the individual pages on the page. As small as they are, it is still quite easy for us to read this content and comprehend. It may not be necessary to read the content of these pages for them to be useful. For instance, if I was only concerned with the changing size and style of the magazine's nameplate, it would not be necessary to read anything more. Once again, Mr. Nesbitt ignores context and the reader's particular information need. However, if I did need to read the content for these illustrations to be useful, I would be stymied because I found it impossible to read the content on these pages at any magnification, despite Mr. Nesbitt's claim otherwise.
The nickel coin is another example of reading the content, 5 cents, but it is used here to provide a frame of reference regarding the size of the page and individual elements.
This a a small format magazine page from National Geographic. On the page are three small pictures. Two, even at this size, are quite recognizable, an elephant, and some deep fried food that probably isn't good for us.
Page 2 of 8
Page 42: This is a small format magazine page from National Geographic. On the page are three small pictures. Two, even at this size, are quite recognizable, an elephant, and some deep fried food that probably isn't good for us. The third image, while rich in content, leaves most of us scratching our heads, unable to decipher the content at any level, without the aid of a text explanation. Mr. Nesbitt demonstrates that the size of a photograph does not always determine its legibility the reader's ability to discern its meaning. Some subjects are more easily conveyed in a small photo than other subjects. With these examples, Mr. Nesbitt proves that small photos of equal size may be legible or illegible. Mr. Nesbitt does not deny that for most news photographs, reduction to thumbnail size reduces the information conveyed.
The third image, while rich in content, leaves most of us scratching our heads, unable to decipher the content at any level, without the aid of a text explanation.
Pages 44-50: This image, which appeared on the AFP site, shows the
affects of resolution and our ability to easily read and comprehend the content. Mr. Nesbitt's discussion of resolution and compression schemes is moot, because all online images are seen at the same resolution (72 dpi) and virtually all photographs are saved using JPG compression. But with these
This example provides a visual demonstration of resolution. Most computer screens have a resolution of between 72 and 96 ppi, pixels per inch.
examples he proves that thumbnails appearing on Google News cannot be used for print because they have too little resolution for print reproduction (which requires 150-300 dpi) and they can't be used elsewhere online at larger size because 72dpi images cannot be enlarged without a loss of resolution and legibility. Pages 45-47: With these examples, Mr. Nesbitt demonstrates that images
This image, which appeared on the AFP site, shows the affects of resolution and our ability to easily read and comprehend the content. Both images are resolved to 8 pixels per inch. The top image was saved in the industry standard, .eps (encapsulated PostScript). The bottom image was then saved as a .jpg/.jpeg (joint photographic experts group). This is not a format but a compression algorithm that has the added benefit of "smoothing" on our screens.
of low resolution, such as 24 ppi to 48 ppi, are far less legible than images of higher resolution, such as 72 dpi. When enlarged, thumbnail images of 72 dpi drop to lower resolution and provide no increase in legibility. So republishing a thumbnail image at larger than original size does not make the content any easier to see.
Page 3 of 8
Pages 48: At 48 ppi, the image becomes clearer and sharper and most viewers can recognize the individual in the picture. Once again, Mr. Nesbitt offers his opinion as fact, with no substatiation for
At 48 ppi, the image becomes clearer and sharper and most viewers can recognize the individual in the picture.
his claim that "most viewers can recognize the individual." I, for one, could not recognize this individual.
Pages 54-55: Regardless, this image can be "read" even at its smallest size of 0.5 inch. The only way the 0.5-inch image can be "read" is if it is seen in direct proximity to the larger images, as presented by Mr. Nesbitt. Seen out of context, the user cannot determine the content, particularly the detail that Mr. Nesbitt cited, in this 0.5-inch image.
However, thumbnail-sized images may be legible when placed in direct proximity to words that provide context for the image. When this occurs at Google News, the context-providing words often come for a source that is different than the image. Pages 56: Even down to the size of most thumbnails, the viewer knows immediately what he or she is looking at. Once again, the only reason the 0.5-inch image can be "read" is that it is seen in direct proximity to the larger images, as presented by Mr. Nesbitt. Seen alone, this 0.5-inch image is illegible.
Pages 57: This image, even to the smallest size signals the viewer:
Series C is what is called a "detail" or "iconic" image. The viewer doesn't need to see the entire body attached to the forearms and hands or the environment in which the image was produced to quickly and easily grasp the content. The content is simple and recognizable to a large portion of the global population.
gymnastics and the rings. Even at 200% magnification, the only way the 0.5-inch image can be "read" is if it is seen in direct proximity to the larger images, as presented by Mr. Nesbitt. Seen alone, this 0.5-inch image is illegible.
This image, even to the smallest size signals the viewer: gymnastics and the rings.
Page 4 of 8
Pages 58-59: In this image the expressions tell us a lot, but it is the body language that provides a wealth of information, even at the smallest size. In the smaller photos, I could not see the body language that Mr. Nesbitt asserts is vital to the meaning of the image. If it is AFP's contention that the content of these thumbnails is as legible as the larger versions, then these examples prove otherwise.
Even if these examples are legible to Mr. Nesbitt, he cannot say conclusively that they are legible to all users without a survey of users to support his contention. I am a trained, experienced and award-winning picture editor, and I could
Series D is perhaps the most complex of the four. Any image containing the human face is read at many levels. Research shows that faces in photos draw viewers' eyes. We read for a sense of the individual, the mood (happy, hostile, sad, neutral, etc.), the situation in which the individuals find themselves, and clues to their identities.
not read the smallest images.
In this case, the two individuals are very well known, Prince Charles and Lady Diana. Once we have established the identity, we search for other cues. In this image the expressions tell us a lot, but it is the body language that provides a wealth of information, even at the smallest size.
Page 61: ...we can discern what the user perceives, assimilates,
It is clear that many of the tenants of print design are still valid, and just as many are not. That someone putting a web page together has the ability to influence how the viewer approaches the screen is not the question. It is rather, are the models now being expressed online working to the best advantage?
and uses, and often why. Without direct inquiry of users, we can't possibly discern why users behave as they do. Eyetracking, for instance, depends upon physiological measurement which tells us where users look, but not why.
Whether it is fixation point, topographic representations, elements and order seen, or movement pattern and motivation (clicks), we can discern what the user perceives, assimilates, and uses, and often why.
A decade ago Dale Peskin, director of The Media Center, called it "another brave new world," and it still is. Much of the data used in this report is open to interpretation. It is the aggregate of this data that is not, and it provides a clear basis for the development of the solutions and conclusions.
Page 66: ...hyperlink underscoring used on the Google News page
Google News uses the same approach as the previous AFP page; however, the material is presented in a denser form. Research has shown that the hyperlink underscoring used on the Google News page inhibits the viewer from reading the blurb. Based on this, viewers will read the headline, look at the image (if there is one), and move on.
inhibits the viewer from reading the blurb. If it is true that the underscoing used by Google inhibits reading, then it contradicts AFP's contention that use of their content at Google News is the same as AFP's.
Thus, headlines, images and blurbs taken from AFP and presented as they are on Google News will result in more viewers just reading the headline and looking at the image, and NOT reading the blurb before they move on.
On the AFP site, the headline is hyperlinked but is not underscored, encouraging more readers to read the blurbs on AFP than on Google News.
In one testing form using two different pages, one with headlines only and the other with headlines and blurbs, the number of "clicks" was low. In all examples the maximum unique click through was 52%. One of the significant findings in this study was that blurbs boosted overall reading across the entire page.
Page 5 of 8
Page 79: One of the significant findings in this study was that blurbs boosted overall reading across the entire page. The study Mr. Nesbitt cites is "Eyetrack III," which was sponsored by The Poynter Institute. The major finding of this report was that words were more important than images to online consumers of news. This conclusion seemed suspect to me when it was first announced with much fanfare in 2004. I took it upon myself to audit the test models used in this study. I learned that many of the pages shown in the test contained no images at all, which accounted for the recorded preference for words over images. As I result, I demonstrated that the major conclusion of this study was probably erroneous. I also found out that this study depended upon a remarkably small number of users to reach its conclusions. In one case, only 14 people were tested. My analysis was published by DESIGN, the journal of the Society for News Design, in its Spring 2004 issue. It appears online at: What follows is an excerpt: For years, newspaper and Web site design has been guided by a central assumption: that readers will more likely read text if it's part of a greater whole, performing its function in concert with photos and other graphic elements. These visuals, as any designer knows, not only relate details and setting and context, but serve as big come-ons to the lackadaisical browser. Poynter said as much in a famed 1990 study it called "Eyes on the News," which found that a newspaper reader's eyes fixed first on photos, and later on the stories they illustrated. So the message I took away from my first reading of that E&P story was either that we've just plain been wrong, or that a reader's habits change dramatically when presented a computer screen instead of a printed page. I took the article upstairs and read it again. The study's subjects, it said, were rigged in Frankensteinian headgear that followed their eye movements, so that the researchers could tell what they looked at, and for how long. Hmmmmmm. But another fact loomed larger on my second pass through the story: Only 67 people participated in the experiments. That hardly seemed a big enough group to yield any firm conclusions about much of anything. I noticed something else, too: Most of the 67 were self-
Only 49% of the participants who clicked a particular headline on the headline and blurbs homepage actually looked at the accompanying blurb. The other 51% never looked at the blurb, just the headline.
Ey e tr ac k III, He ad lines a nd blurbs
Page 6 of 8
professed newsaholics who joined the study by answering online newspaper ads who, in other words, were already habitual newspaper Web site readers, rather than members of the Internetusing general public. As I mulled this over, my shock began to give way to skepticism. Before I paid the news any mind, I decided, I had to read the study itself. I found it on the Poynter Web site. Of the test's 67 readers, I learned, only 14 were examined for their behavior in front-page attention. These 14 saw fewer than seven pages that contained a single photo. At this point, my skepticism was replaced by outright disappointment. How on earth had a study of too few subjects, of too narrow a persuasion, reading too little, made it into public view? Clicking through the report just brought more doubt. Of the three sample pages published at the Poynter site, only two had photos and they, just a single photo each. Pretty tough to test a reader on his preferences for text versus photos if he's got nothing but text to look at, it seemed to me. On top of that, neither of the sample page photos contained the image of a person, which may have been more appealing than what was offered. I was tempted to toss the magazine in the trash, but I was too irritated to do much but slump in my chair and mutter to myself. My sylvan morning had been wrecked. I was far too distracted to work. I took a walk around the block. Upon my return it occurred to me that the Poynter/Stanford study might have value that I hadn't initially recognized. True, I believed its research was fatally flawed. Convincing me of its conclusions would take a proper study, involving far more subjects, from a far broader bandwidth of humanity, spending more time at the keyboard. To my knowledge, no one uses the findings of Eyetrack III to guide the design of news Web sites. Poynter has announced plans to conduct a new study, in part to correct the deficiencies of Eyetrack III. Even more remarkable is this statement from the Eyetrack III managers: ...a quick comment on what this study is and is not: It is a preliminary study of several dozen people conducted in San Francisco. It is not an exhaustive exploration that we can extrapolate to the larger population. With this statement, the managers of Eyetrack III seem to be saying that their findings have limited applicabilty, if any, for practical purposes.
Eye tracking measures a large number of user data. Individual tracking sessions provide a visual map of the user's entry point on the page, the path of the eye (saccades), fixation points, and exit point. Fixation maps deal primarily with what elements on a page the user stops and spends time comprehending. Heat maps are a combination of individual tracking sessions, showing the percentage of users that fixate on certain parts of the page. These percentages are broken down by color with red/orange indicating that almost all participants halted their gaze, yellow indicating the more that half halted their gaze and blue/green representing less than half. In addition, many heat maps of Web pages indicate the position and number of "clicks" made by the participants.
Page 7 of 8
Page 83: Eye tracking measures a large number of user data. Individual tracking sessions provide a visual map of the user's entry point on the page, the path of the eye (saccades), fixation points, and exit point. Mr. Nesbitt conveniently leaves out what this research does not tell us: What the user understands and why they choose to look at one element over another. These two criteria are far more meaningful than any other.
Page 87: This research supports AFP's contention that the use of
This research supports AFP's contention that the use of their material by Google in Google News is the same use as AFP. In addition, this research shows that in most cases, the AFP material, as a self-contained unit, is more than enough information for the user.
their material by Google in Google News is the same use as AFP. Mr. Nesbitt offers a tremendous amount of information, some erroneous, some contradictory, some not supported by common sense, but none that asks users, "Do you use Google News the same way you use AFP?" Furthermore, Mr. Nesbitt does not support his contention that use of material by Google in Google News is the same as AFP. In fact, they are fundamentally different because AFP serves an audience of paying clients,
The research indicates that the number of "click throughs" to the full length story or article seldom reaches 50% of the users, and is often far lower.
while Google News serves the general public for free and for a different purpose.
Page 87: In addition, this research shows that in most cases, the AFP material, as a self-contained unit, is more than enough information for the user. Mr. Nesbitt offers no research that shows that in most cases, the AFP material as a self-contained unit, is more than enough for the user, because he offers no test results of the AFP material with users.
Page 88: In the following two examples from Eyetrack III, the two
In the following two examples from Eyetrack III, the two pages differ in the way that the information is presented to the viewer. In the first example, the headline and the blurb text is small, as are the topic titles.
pages differ in the way that the information is presented to the viewer. These pages differ in a more fundamental way: Neither of these pages accurately reflect the appearance of typical news sites in several important ways: On most news sites, the largest graphic element is an ad, which in many cases is animated; Eyetrack III test sites did not include such advertising. Most news sites contain multiple ads in the first screen; Eyetrack III pages did not. Most news sites have a highly developed navigation system; Eyetrack III pages did not. This lack of fidelity with
In the second example the headline, blurb and title text is larger. Both pages have headlines only at top, and headlines and blurbs at the bottom of the screen.
In both cases, the headlines and blurbs are presented in a similar format to AFP and Google News.
Page 8 of 8
actual news sites, among other things, casts serious doubt on all the findings of Eyetrack III.
Pages 90-94: There are two primary reasons for the low number of click-throughs. One, the information needs of the viewer were already satisfied; and two, the topic or event was of little or no interest to the viewer. Mr. Nesbitt admits that a lack of perceived relevance may account for the lack of click-throughs. Now consider the actual content of some of the headlines:
·Names stuck? It's All In Your Head ·Has Stanley Williams Left the Gang? ·Soccer: The Magic of Man U None of these headlines provide a clear description of the story which may account for the lack of click-throughs. So there is yet another reason
As the heatmaps show, there is a high concentration of user fixation on the headlines, more so on the smaller type version. We also see more "click throughs" in the smaller type version. Still, in this example, only 24% of the participants clicked in the upper right area of headline news, regional, and sport. In the larger-type version, only 9% of the participants clicked through to the full story.
beyond user satisfaction that may account for a lack of click-throughs. Clearly, user satisfaction has not been proven to be the only explanation, which contradicts AFP's contention. Furthermore, when a mere headline or blurb satisfies them, it may be because only basic facts were of interest.
In both cases, almost 100% viewed the headlines but fewer than one quarter clicked for additional information.
Page 95: In the second example, headlines and blurbs, the highest
The next two examples were used in the study to track the differences between a page with headlines only and headlines with blurbs. As can be seen in the first example, headlines only, there was a high concentration of user fixation on the headlines, especially in the first screen, and only 33% unique click throughs were recorded.
concentration of fixations were on the sex offender story. With this example Mr. Nesbitt proves my point: it's the content that drives usage, (in this case, the words "sex offender") more so than format.
In the second example, headlines and blurbs, the highest concentration of fixations were on the sex offender story. However, in the same geographic area of the page, there were only 21% unique click throughs reported. With more information supplied by the blurb, fewer readers felt the need to go to the full story for additional information.
Disclaimer: Justia Dockets & Filings provides public litigation records from the federal appellate and district courts. These filings and docket sheets should not be considered findings of fact or liability, nor do they necessarily reflect the view of Justia.
Why Is My Information Online?