Jump to content

Wikipedia talk:Manual of Style/Accessibility/Alternative text for images/Archive 4

Page contents not supported in other languages.
From Wikipedia, the free encyclopedia
Archive 1Archive 2Archive 3Archive 4Archive 5Archive 6

Help with some alt text

I've been alt-texting Bird and could use some thoughts on two of the images remaining. Image:Birdmorphology.svg is a morphological map of a bird which ties into the description below. I don't know how to do it while keeping it brief. Similarly File:Neoaves Alternative Cladogram.png strikes me as a very hard one to image alt. Thoughts? Sabine's Sunbird talk 03:00, 8 March 2010 (UTC)

  • Image:Birdmorphology.svg could perhaps be done with something like the following: "Diagram of a bird standing with wings folded, with numbers identifying points on the bird. Numbers 1 through 11 march along or near the top of the bird from its beak to its tail; 12 through 16 march down the bird's rear to its feet; 17 through 22 then march upwards along the bird's front, ending just under its beak."
  • File:Neoaves Alternative Cladogram.png is pretty much useless as an image: one cannot make out anything in it at 300px. I suggest removing it (independently of the alt text issue). Or, if you keep it, it needs to be at least 500px to be useful, but that's a tad large. Perhaps it'd be better to redo it as text, or a subset of it using {{clade}}? Possibly this could also involve wikilinking to the image, using "[[:File:Neoaves Alternative Cladogram.png|very brief explanation]]". The file page could then contain a complete textual description. If you really want the image as-is, with alt text, I suppose you could summarize it along the lines suggested in the previous bullet.
Eubulides (talk) 04:35, 8 March 2010 (UTC)


The lead example

"alt=In the sky flies a red flag with a white cross whose vertical bar is shifted toward the flagpole. |The oldest national flag design still in use is Denmark's 13th-century Dannebrog."

I recommend replacing this text with:

"alt=A flag flies from a flagpole against the sky. The flag is red, divided in four by a white cross, the vertical bar of which is shifted toward the flagpole. |The oldest national flag design still in use is Denmark's 13th-century Dannebrog."

Reasons:

  • "In the sky flies a flag" is convoluted. Let's have "A flag flies" with the subject first in the description, and before the predicate rather than after it.
  • "a cross whose vertical bar" is grammatically incorrect. The cross is not a person.
  • The addition "divided in four by a white cross" makes sure that the form of this cross cannot be mistaken from the one on the Swiss flag.

Amandajm (talk) 13:00, 8 March 2010 (UTC)

Thanks, fixed. Eubulides (talk) 00:50, 11 March 2010 (UTC)

Changes

I have just made a number of changes here, without consensus I'm afraid. There seems to be a fairly clear consensus that it is the needs of the visually impaired that we are catering to, yet at the same time a wish avoid all sorts of things that make a blind person's listening easier.

  • I obviously need to repeat again that the blind person needs the subject identified as soon as possible within the alt text. It doesn't matter whether the name "Napoleon" is repeated, as long as the blind person knows from the start that the man with a brutus hair cut and his hand in his waistcoat is in fact Napoleon and not Edward VII at the age of fifteen at a fancy dress ball. Repetiton is the least problem. Although to some wikipedeans this apparent lack of efficiency is an anathema.

Some editor in response to me previous said that a sighted reader looked a a picture of a person and then read the caption, not knowing the identity of the mustachioed man until they read his name. What you need pointing out is that the glance at the face and the scanning of the caption happens in an instant. The sighted person doesn't look at the image in detail before readin. They simply take it in. For a blind listener, there is a time lapse, and a waiting to discover, and a possible wish, "Well, now that I know that it's Ned Kelly they are reading about, how do get the description back?"


  • Whether or not to state the medium. After having a blind parent, working with blind children and conducting tours/writing educational material for the blind, my advice is that it is often useful to state the medium.

For example: There is a very big difference between a studio photograph, a press photo, a snapshot and a mugshot. The context in which the media was created dictates the type of content. Your sighted viewer processes this information, on looking at the image. Your blind viewer is entitled to it as well.

  • While it may not always be necessary or desirable to state that an image is a painting or whatever, this also provides context. It can be important for both blind and sighted readers to know the date of an image. For example, a well meaning editor recently deleted the 16th century chalk portrait of Leonardo because no-one knows for sure whether it truly is a self-portrait. This person replaced it with a 19th century encyclopedic engraving based on the same image because its text said that it was Leonardo. Obviously the much-debated image has infinitely better qualifications than a derivative made four hundred years later.

In this line, a 19th century French painting of Joan of Arc is not Joan of Arc, and it's also not a portrait of Joan of Arc. If she and her armour are described, then the 19th century context needs to be clear. This stands for every historic character romantically depicted in paintings.

Need coffee! Amandajm (talk) 14:56, 8 March 2010 (UTC)

Please see #Identification and repetition below. Eubulides (talk) 00:15, 9 March 2010 (UTC)

Identification and repetition

  • "the blind person needs the subject identified as soon as possible within the alt text" Yes, but that identification need not be the name of the subject; describing the subject not only suffices, but is the point of the alt text.
  • "The sighted person doesn't look at the image in detail before readin. They simply take it in." Nobody can understand the image and the caption instantaneously. Even sighted readers need time to look at the image, and to read the caption. It's true that a fast sighted reader can do it faster with an image than a fast blind reader can do it with the image's alt text, but that's true for the textual part of an article as well: there's nothing unique about alt text here.
  • "For a blind listener, there is a time lapse, and a waiting to discover, and a possible wish, 'Well, now that I know that it's Ned Kelly they are reading about, how do get the description back?'" This scenario is implausible for the Ned Kelly article, because a reader listening to the article will hear something like this:
    "Ned Kelly. from Wikipedia, the free encyclopedia. Jump to: link navigation, link search. For other uses, see link Ned Kelly (disambiguation). Link This article is semi-protected. Edward "Ned" Kelly link Head of a young man with a long, untrimmed beard..."
  • where I've underlined the start of the alt text. Here, the visually impaired reader will already know that it's Ned Kelly that's being described. This is the normal case for articles about people.
  • "it is often useful to state the medium." Yes, it often is, particularly in the Leonardo example that you give, where the focus is on an artist, and the medium is an important part of the art being illustrated. However, for most images in Wikipedia it's better to not state the obvious "photo of" or "image of" or "this image shows that...". It might be useful to state the exceptional, art-oriented cases more clearly, but the main point should be retained. The overall point has been in this guideline for many years, and with widespread consensus. See, for example this 2008 version, which says, "Alternative text describes the image, not its file or format. Thus, you shouldn't mention ... the fact that it is a substitute for an image: ... Bad: 'This image shows that...'".
  • "If she and her armour are described, then the 19th century context needs to be clear." It's fine to explain context, but it has to be done by describing the image itself, not by using phrases like "19th century painting". That would be giving information to visually impaired readers that sighted readers typically will not know. Any information such as a painting's date should be placed in the caption, so that sighted readers are given the information as well. Alt text should not contain secrets hidden from sighted readers.
  • The changes that you installed had some good ideas, but had several problems as well.
    • Insertion of proper names into alt text is unwise, unless it's an icon such as Napoleon or is a person or other entity already described earlier in the article, for reasons discussed above (also discussed in WP:ALT #Proper names). Certainly it was wrong to remove the recommendation that non-iconic figures like Putin should be named in the caption or other text: again, alt text should not contain secrets hidden from sighted readers.
    • The examples are meant to be actual examples from Wikipedia, as much as possible. When we change the alt text here, we should also change the alt text in the actual pages. In the past, when I've done that, the editors of those pages have changed the alt text back. I'll go and change those other pages, to keep them in sync with what has been done here so far; I just thought I'd mention it now as a constraint.
    • The removal of WP:ALT#Repetition is incorrect. The requirement that alt text should not duplicate what's in the caption has been in this guideline for quite some time; see, for example the same 2008 version, which says, "The alt text description should not duplicate information already present in the caption or the main text of the article." And the idea that the alt text shouldn't repeat the caption is not just a Wikipedia thing: it's standard practice web-wide. See the W3C guideline on adjacent image and text links, which explicitly warns against duplicating information in alt text and adjacent text. It wastes time to tell a visually impaired reader the same thing twice, once in alt text and then again (repeated) in the caption. We've had comments from a blind Wikipedia editor that alt text should be brief; see, for example, Wikipedia talk:Featured article candidates/archive39, in which Graham87 says "I don't want information overload, just a rough mental idea of what's in an image" and "I find alt text helpful when it describes properties of an image that are easy to discern for sighted people but aren't appropriate for a caption": this underscores the basic principles that alt text should (1) be brief and (2) not repeat the caption. It may well be that WP:ALT#Repetition can be improved by rewording it, but removing it is not an improvement.
  • To attempt to work around these problems, I installed the following further changes. They keep most of the suggestions for improving the examples, but omit the controversial bits about proper names and repetition. Further comments are welcome.

Eubulides (talk) 00:15, 9 March 2010 (UTC)

You seemed to have missed the point above about the order. If the caption was above the image then your reply would be sensible but the caption is below. Maybe somehow the caption is read first? if so please expand on that so we can understand. Otherwise I feel that having read the previous post and the reply is seems the previous poster makes a good case for overturning old ideas and using repetition and possibly for proper names to be included in the alt text also. Regards, SunCreator (talk) 05:05, 9 March 2010 (UTC)
How did I miss the point? I transcribed the text from Ned Kelly accurately, and as that example indicates, there's absolutely no need for its alt text to mention Kelly's name, as it's already obvious from context. I was responding to the example given in the preceding comment, but if you prefer a different example, please look at the first image in WP:ALT. There's no need for its alt text to say "Dannebrog" (the name of the flag in question). Generally speaking proper names like that ought to be in the caption (because the reader can't be expected to know them) and shouldn't be in the alt text (for two reasons: first, they don't explain the image's appearance to the reader, and second, they repeat the caption). Eubulides (talk) 05:25, 9 March 2010 (UTC)
Thank you Eubulides. I am beginning to see where the wires are crossed. To repeat your part
For Ned Kelly listening to the article will hear something like this:
"Ned Kelly. from Wikipedia, the free encyclopedia. Jump to: link navigation, link search. For other uses, see link Ned Kelly (disambiguation). Link This article is semi-protected. Edward "Ned" Kelly link Head of a young man with a long, untrimmed beard..."
The above works because you've been selective in an example at the top of an article where the photo is of the person of the article name.
Next however consider World Chess Championship
"World Chess Championship. from Wikipedia, the free encyclopedia. Jump to: link navigation, link search. Link This article is semi-protected. World Chess Championship link A man sat looking down at a chess table on with his arms crossed. He has the white pieces, to his side is the flag of India. In the background and on the chess clock you can read the words 'Chess classic'..."
Here you don't have the context of a person for the article. So it doesn't work and you are better off to start by say "Current World Chess Champion Viswanathan Anand sat looking down at a chess table..."
Next consider that most images are not the top image in an article introduced by the article name. Most images are in section further down the article where there is no context about the person. For example see Personal_relationships_of_Michael_Jackson#Early_sexual_and_emotional_experiences here we have Early sexual and emotional experiences. Strip clubs. An older man wearing a suite and has an earring etc but I'd rather get to the point and know who it is. Regards, SunCreator (talk) 00:07, 11 March 2010 (UTC)
  • The alt text of World Chess Championship's lead image should not name Anand. First, that image is primarily of a man playing chess: the man's identity is a trivial detail of the image's visual appearance in that context (much as the image contains other trivial details, like the color of the wall). Second, giving the name in the alt text does not help the typical reader, who can't be expected to know what Anand looks like. Third, the name appears in the caption (for identification), and the article should not burden the visually impaired reader with unnecessary WP:ALT#Repetition. Other examples, such as the one you mention for Michael Jackson, are similar: people's names are not necessary or helpful in the alt text itself, even if they are a helpful part of the context provided by the caption. The most important aspect of that image in the Michael Jackson page is the glowering face and the flashy clothes, not the man's name.
  • There seems to be a misapprehension here about how screen readers work. People don't listen to the screen reader passively, waiting for it to read the article from beginning to end. They flip around from one spot to the next, rapidly, as a sighted reader does. (The whole process takes longer, but the flipping part is still there.) It's easy and common for someone reading an image's alt text, and who wants to know what's in the caption, to tab forward to the caption, and then tab backward to the alt text (or forward to the next section, if they prefer). So, a reader who really wants to know who that glowering figure is, and can't stand to wait for the caption, can go listen to the caption, and come back to the alt text.
Eubulides (talk) 00:50, 11 March 2010 (UTC)
Firstly sorry for slow reply but it is almost impossible for me to read and edit this conversation on an iPhone. Can we split up the text and sections and also indent in a readable way - Or perhaps easier just start new sections with one issue at a time below? Useful stuff about the screen reader; Eubulides I hand't realised that. Your Ted Nelly example and the original part

about listening was making me think of listening like you can listen to article text, I guess that I was not correct in that thinking. The Repetition reply you gave above leaves be totally baffled, I see no reason is given for this avoidance of repeat. I'd like to discuss that seperately, it seems a matter of priority for me and repeating text seems low prioirty and the least of our concerns. Regards, SunCreator (talk) 01:52, 16 March 2010 (UTC)

Agree with Eubulides. It seems fairly simple to me: alt text describes the image first (the visual aspects), and the caption then interprets it (names, setting, importance etc.). Dabomb87 (talk) 03:25, 10 March 2010 (UTC)`
What you have written there, Dabomb87, is the problem. However, you are wrong in saying that the caption interprets by giving name and setting (let's include date). These are not matters of interpretation. These are the facts that are known about, and provided with almost every image on Commons. (If they are not provided, then they should be.) This is not interpretation. Interpretation may be the significance of the image. The significance is not alt description stuff. It is caption stuff. The name or identity of the person, object or place is essential information. (unless of course the picture is of a generic "person").
At the risk of seeming repetitious,
  • the blind person needs to know, from the start, what the subject is that is being described. A sighted person can take in an image, and a name beneath it within a second, or possible three if they are a slow reader.
Let us take for example, an article which mentions the Mitford sisters, the signatories to the Declaration of Independence or the Members of The War Cabinet. (The examples, like my Ned Kelly example, I'm sorry to say, are off the cuff).
We have five photos of five sisters. Five descriptions. If the writer is good at describing, then each description is different.....but the blind reader has no idea which one is being described until after the description. To insist that this is appropriate seems to me, (personally) to be absolute stupidity, but then, that opinion is simply based on the training and conducting of tours for blind people. Call it "Original Research" or "Personal Opinion" if you like. I call it "Experience".
  • It's fine to explain context, but it has to be done by describing the image itself, rather than giving the date....

How, please, how does one describe "the context" of a picture that has been done hundreds of years beyond the lifetime of the subject without giving the date?

I am a very good describer, and I could describe the Joan of Arc painting to which I have referred in such detail than an expert in the field of French Classical painting would say "I believe that is a French Classical painting that you are describing!".... but the average person would not. A date is by far the easiest way to convey the fact (to a blind person) that the image is not from the lifetime of the subject and does not represent a true portrait. On the other hand, although a sighted viewer might not know whether the picture is 18th, 19th or 20th century, they will almost certainly know immediately that it is not medieval, but dates from many years after the life of the subject. Most people's general knowledge would extend this far.

Similarly, a person of reasonable intellect will also know immediately that the print of Hadrian and his lover Antinous is not an ancient Roman original, while a glance will tell them that the damaged portrait of Septimius Severus, his wife and two horrible sons probably is. Your blind listener has the right to know this as well. If the date of the image is included in the caption, then it doesn't need to be in the alt text. It is also common courtesy to acknowledge the creator of an artwork (in the caption) whenever one is used.

Because users of computers are very familiar with photographic images, they can usually tell at a glance whether a picture is "formal" or a "snapshot" (unless a posed picture has been skilfully contrived). This is important contextual information that the sighted person doesn't need to be told, but which is useful to a blind person.

  • There is a statement somewhere that colour doesn't need to be described. In some cases this may be so, but please keep in mind that a great number of blind people have previously been sighted, or have residual sight and can see a little colour.
  • The amount of information required in an alt description differs from subject to subject.

A description such as "Dark-haired clean-shave caucasian man in a dark suit and white shirt" is such a useless description that you might as well not even bother writing it. Even if you add a detail like "balding", "smiling", "with glasses" etc, it adds nothing. "Two middle aged men in dark suits at a table with two younger men in dark suit" also adds nothing of value. If there is going to be a description, it needs to lock into that part of the subject that really requires describing.

  • This descriptive process is often a very subjective thing. A sighted person reads faces. A blind person reads tone of voice, involuntary noises, small movements such as shuffling, and smell. But they don't get any of this from a photo that they can't see. So the describer needs to describe what they can see. It is easier to say a man is "scowling' than to say his brows are pulled together and his mouth is turned down. Even a blind person who has never seen a "scowl" has still heard the word and knows what sentiment it carries. This may sound too interpretative for Wiki, but nearly all sighted people have the ability to recognise visual cues like this. Telling the blind person that the person is scowling, looks content, looks sly, looks anxious etc is telling the real story behind the image that is there for every sighted person to see. Yes, I know that you can dig up Wiki quotes to counter what I am writing here. The point is, do you want to write effective alt texts, or not?
  • The amount of information required differs from subject to subject. In many cases brevity is desirable. In other cases a full description may be advantageous, because the image itself conveys useful information. If it is an artwork in an art article, for example, then a full description may be essential for some, or perhaps many, of the images, depending on the article.
  • The amount of description desired varies from person to person. Graham87 doesn't want information overload. Many other blind people love hearing description of what others can see. Just as some people prefer to read maths textbooks and others prefer to read poetry. I don't know what Graham87's particular subject is, but I do know that my blind relative has always liked things described in full detail. Once again, I think that the subject matter has a bearing on this.
  • I have to ask, seriously, if one is not supposed to say that the "man in a dark suit" is Putin or Obama or Dickens, then how far to we carry this non-repetition? When we describe a big stone object beside a river, are we allowed to tell them straight away that it is the Tower of London, or do we keep them in suspense during the description, until they have eliminated the Statue of Liberty, the temple of Abu Simbel and the south east pylon of the Sydney Harbour Bridge?
  • Does this rule of non-repetition only apply to Proper Nouns or ought we extend it to other nouns as well? Wandering Albatross, for example: do we call it an albatross in the alt text or do we say "big white bird"? (But if we cannot say that a humungously large spiral staircase is "vast" then I'm sure that "big" is also too descriptive. Better just keep it down to "white bird".) Are we permitted to say that a polar bear is a polar bear, or must we first describe it as a bear with white fur? Is it alright to tell the blind reader that the flower in the picture is a rose, or do we just call it a pink flower and make them wait to find out? And as for that flapping stuff on the flagpole, do we call it by the generic name of "flag" and then repeat that word in the caption, or should we say first that it is a piece of red cloth?

Amandajm (talk) 13:57, 10 March 2010 (UTC)

Common sense should apply. Write for the typical reader. One can assume that a typical reader knows what a flag is, so it's better to say "flag" than "colored and rectangular piece of cloth flapping in the wind". On the other hand, one cannot assume that a typical reader knows what Putin looks like, or can tell the difference between an 18th century and a 19th century portrait of Seneca, so these details should not be in alt text. To take your example of the Mitford sisters, please look at File:Mitford sisters.jpg in the context of Mitford family #Mitford sisters. The sighted person glancing at this image gets a group impression, not individual impressions, so the alt text need not (and should not) go into a detailed description of each woman individually. I just now added the following alt text for that image:
"Nearly full length group portrait of five well-dressed women standing in a field. Their ages range from roughly 20 to 30; their hair is cut short of the shoulders in elegant 1930s to 1950s styles; four of the five wear skirts down just below the knee, and one a longer coat. Two wear pearls."
and this is reasonable alt text. There's no need to name the individuals here: the names are irrelevant to the visual depiction, and alt text is supposed to be an alternative to the visual depiction, not an explanation or identification of it. Eubulides (talk) 17:54, 10 March 2010 (UTC)
In this instance I was using a hypothetical case (as I pointed out) in which there might be five individual portraits of females, all described differently. Mitford sister, Bronte sisters, Romanoff sisters....OK!
Likewise, five individual portraits of the Jackson five, five indiv. pics of Acka Dacka ....Amandajm (talk) 23:14, 10 March 2010 (UTC)
I'd think that such examples (we're talking about a gallery, right?) are pretty rare. There's nothing like that in Jackson Five (sorry, I don't know what "Acka Dacka" is). Can you supply a specific example that's already in Wikipedia? Eubulides (talk) 00:50, 11 March 2010 (UTC)
As for some of your other points:
  • It's fine to say whether a portrait is formal or a snapshot, particularly if this is an aspect of the image that stands out.
  • "It is easier to say a man is "scowling' than to say his brows are pulled together" Yes, that's absolutely right, and it's the point of the Neutrality section.
  • "keep in mind that a great number of blind people have previously been sighted, or have residual sight and can see a little colour" Yes, that's quite right. The point about not saying color is to avoid nearly-useless noise phrases like "color photograph of" at the start of alt text. If the point of the image is to illustrate a red balloon, then obviously the alt text should say "red".
  • "How, please, how does one describe "the context" of a picture that has been done hundreds of years beyond the lifetime of the subject without giving the date?" If that context is obvious from the picture, then it ought to be easy: just describe the gist of the picture. But one should not give more information than is obvious from the picture itself. For example, it is not obvious to a non-expert viewer that the image File:Ingres coronation charles vii.jpg is a 19th-century painting (it could easily be from an earlier or later century), so the alt text for that image shouldn't say "19th-century".
Eubulides (talk) 17:54, 10 March 2010 (UTC)
Maybe it could be from an earlier or later date!
But it works like this: If I describe that picture well enough to give it a context, then some people (a few people) will know purely from the description that it is 19th century. That number is far far less than the number of sighted people who will recognise it immediately as 19th century by appearance alone.
Conversely, if I tell the non-sighted reader that the painting is 19th century, then a everyone who has been sighted, and has had any interest in art whatsoever, will almost certainly know that the painting is detailed, realistic and probably rather dramatic. The reason for this is that although "detailed, realistic and rather dramatic" could mean a large number of things, on the other hand "19th century painting" almost always means these things, (unless of course, one states that the picture is "Impressionist" which conjours a different image, even in the extremely ignorant.)
However my preferred option is to state in the alt text that it is a "painting", and state in the caption both the date and author.
A few works like the "Mona Lisa" are iconic. Only seven artworks fit into that category: Leonardo's Last Supper (by that name), Mona Lisa, Creation of Adam, Michelangelo's David (by that name), Botticelli's Birth of Venus, Venus de Milo and Leonardo's Vitruvian Man. If you mention these, then the average person's general knowledge knows that you are referring to a very great and famous artwork, whether they can picture it in their minds or not. Don't presume too much ignorance.

Priorities

  • It seems to me that the priorities being expressed here and on the MOS page are wrong. It really worries me that you people are so locked in to avoiding redundancy, and avoiding giving the blind reader some precise detail that the sighted reader may not have. These are the least important matters. You seem to forget that the sighted reader can simply click on the image and, provided the uploader has done their job, get a fuller description and even view the details enlarged. The blind listener doesn't have this easy option.
  • I also want to comment that many images don't need a detailed physical description at all. Take the Abbey Road terrace, for example. If I was using this as an example of a London terrace house on a page about architecture, then it would require precise descripton. On a page about a Beatle's album it is sufficient to say that it is a four story, red brick terrace house. The number of panes in the windows and the position of the door and whether it has an attic don't matter in the context of the article in the slightest.
  • While brevity is desirable, the addition of a single word that informs the enlightened can be advantageous. A single word can speak volumes to those people who have an interest. For example, to the description of that house, I would add the single word "Georgian" because that word illuminates both the date and the style to many thousands of people, if not to others. To everyone who is interested in such things, that single word "Georgian" tells them that the house is neat, flat-fronted, un-fancy, and has large sash-windows. "Victorian" terrace conjours an entirely different image. Are these descriptions only for the expert few? No. They are for every person who has ever flipped through the real estate advertisements. If the person is not sufficiently interested to know the difference between a Georgian terrace and a Victorian terrace, then there is only one little word to skip over, rather than a lengthy description.
I don't recommend replacing descriptions with technical terms understood only by a few. But I do recommend the inclusion of those short terms that speak volumes to many.
  • It worries me a great deal that the lead picture of the article instructing editors on how to write alt text and beginning with "Sky with flag...." rather than "Flag...." should have stood so long as the prime example of good alt text writing. I have no idea who wrote it or who selected it. But, frankly, I don't want to have to argue with either of those people. At the risk of causing offense, I suggest that they should pull their heads in.

Amandajm (talk) 23:14, 10 March 2010 (UTC)

I agree with you Amandajm. Regards, SunCreator (talk) 00:15, 11 March 2010 (UTC)

I also agree with Amandajm, as follows:

  • "my preferred option is to state in the alt text that it is a "painting", and state in the caption both the date and author." That's fine, for a page discussing art. In such an environment one normally would expect alt text for a painting to say that it's a painting. I added something along those lines.
  • "Don't presume too much ignorance." That's also fine, and WP:ALT #Proper names already says it's OK to give names of iconic images.
  • "many images don't need a detailed physical description at all" Yes quite true. WP:ALT#Brevity is supposed to cover this issue, but I see now that it wasn't stated clearly; I added that.
  • "I don't recommend replacing descriptions with technical terms understood only by a few. But I do recommend the inclusion of those short terms that speak volumes to many." This is also fine: one must write alt text for the expected audience.
  • 'It worries me a great deal that the lead picture of the article instructing editors on how to write alt text and beginning with "Sky with flag...." rather than "Flag...."'. Thanks for reminding me about that; I followed up in #The lead example above.

I see one point where we disagree: "the sighted reader can simply click on the image and, provided the uploader has done their job, get a fuller description and even view the details enlarged. The blind listener doesn't have this easy option." No, the blind listener can follow the link to the image, just as the sighted reader can. It's just as easy for the blind reader to follow this link, as it is to follow any other link. It's true that the blind reader can't view a larger image, but the blind reader can get a fuller description. WP:ALT#Brevity talks about this issue, and recommends putting a longer description in the file page, precisely for this reason. Eubulides (talk) 00:50, 11 March 2010 (UTC)

That's very good! I'll remember that when writing file pages and try include fuller historic details, etc. Amandajm (talk) 01:43, 13 March 2010 (UTC)

Alt-poetry

I did not realize that article space has a place for subjective, fanciful, poetic etc. fiction ... It's right here!

  • "deep-set eyes gaze solemnly over the viewer's shoulder ..."
  • "serious and dignified ..."
  • "with a haunted expression, one hand raised ..."

No, it's not Hitler in Nuremberg, it's about three dead white women. And "large ship" (sic) is not Batillus but the diminutive SMS Helgoland. At 167 meters length it was not larger than modern destroyers.

Shouldn't alt-text be governed by general style, RS and BLP rules? "Living person X has 'ill-kempt, wavy hair, a high forehead, and deep-set eyes with a wary, watchful expression ... '" is unacceptable in article space, what makes it good for alt-text? NVO (talk) 09:26, 4 March 2010 (UTC)

Please see WP:ALT #Neutrality. Eubulides (talk) 16:11, 4 March 2010 (UTC)
My point exactly. "Neutrality" in one paragraph, "solemn haunted dignity" elsewhere. NVO (talk) 16:31, 4 March 2010 (UTC)
Non-verbal communication requires interpretation of gestures/body language/facial expressions by a viewer. If you want to give a better idea of an image, human interpretation will be required else the plain matter-of-fact description makes it harder to get a good grasp of the image for a listener. AshLin (talk) 06:29, 19 March 2010 (UTC)

Purely decorative?

Are the images in List of Olympic medalists in curling#Athlete medal leaders purely decorative, or do they require alt text as well? Dabomb87 (talk) 23:51, 8 March 2010 (UTC)

They need alt text, as they provide useful visual information that is not in the adjacent text. Eubulides (talk) 00:17, 9 March 2010 (UTC)
Thanks, I added alt text. Dabomb87 (talk) 00:33, 9 March 2010 (UTC)

Promotion of alt text

Can we have some sort of promotion. It seems use of alt text is almost none existant on most articles. Yesterdays featured article didn't have any alt text a few hours before hitting the main page; the same with tomorrows featured article and both would of sailed on without it if I hadn't raised the issue. Discussing the finer points of alt text seems irrelevant if it's going to be largely ignored anyway. Regards, SunCreator (talk) 04:46, 9 March 2010 (UTC)

We've promoted it in featured articles by bringing it up consistently in Wikipedia:Featured article candidates and in Wikipedia:Featured article review, so that new (or newly-reviewed) featured articles all have alt text nowadays. It'd be quite a bit of work to convert the other featured articles (not to mention the rest of Wikipedia), but further promotion of it would be welcome. This topic came up recently in Wikipedia talk:Featured article candidates #Alt text in FA images and you might want to check out that thread as well. Eubulides (talk) 05:32, 9 March 2010 (UTC)

Obscured text

I've been updating images in Talyllyn Railway, and tagged this image with the alt caption "A small green tank steam locomotive of unusual design, number 13, stands in the museum. The power to the wheels comes from a crank-shaft at the upper rear of the locomotive, which is transmitted to the rear wheel via a vertical connecting rod on the right hand side. The rear wheels are connected to the front wheels by a conventional horizontal connecting rod. A plaque on the front of the locomotive reads WILLIAM SPENCE S. GEOCHECANS PATENT, with a final line obscured by a hand rail rising from the front buffer beam."

I've recently been back to the museum, and noted what the obscured text was. Is it better to leave the alt description as is, or update it with the missing line of text (which would be more informative, but a less accurate description of the photo)? —  Tivedshambo  (t/c(logged on as Pek) 15:26, 10 March 2010 (UTC)

The solution is to remove the mention of the plaque, its text and the handrail (which latter adds no value to the description) from the alt description and write the entire text in the caption so that all readers have the benefit of it. Amandajm (talk) 23:22, 10 March 2010 (UTC)
Also, leave out the words "on the right hand side". This implies that there is only one connecting rod, and that it is located on the right. If it simply states that the power is transmitted to the rear wheel via a connecting rod, then your listener will presume symmetry. Which is exactly what the sighted viewer presumes, when looking at the photo. Amandajm (talk) 23:30, 10 March 2010 (UTC)

Policies issues

I've been WP:BOLD and adding the following:

Policy issues

Occasionally the descriptive nature of alternative text creates issues of keeping to policies such as no original research and neutral point of view. If it is not possible to write a descriptive alternative text that clearly passing policies then the alternative text should revert to being a suitable identifying text as used in caption text.

You can understand the requirement by reading the entire text from this FA discussion. Regards, SunCreator (talk) 18:04, 12 March 2010 (UTC)

That topic is already covered in the WP:ALT #Constraints section. There is certainly no consensus for saying that alt text should just be a caption, so I've reverted the addition of this section. I suggest proposing wording changes to the Constraints section instead. Is there anything in that section that's missing, or where the wording needs improvement? Eubulides (talk) 19:10, 12 March 2010 (UTC)
Agree with removing the above. There is no point in repeating caption text. Colin°Talk 20:16, 12 March 2010 (UTC)
That being said, the idea that the alt text should in some cases be simply a placeholder for adjacent text is worth briefly mentioning in this section too, with a wikilink to Placeholders for more details. How about this edit? Eubulides (talk) 20:18, 12 March 2010 (UTC)
I don't see that avoiding repeating text should take priority over compliance with core policies. If repeating text on occasions avoids WP:OR and WP:POV then I welcome it. Regards, SunCreator (talk) 23:12, 16 March 2010 (UTC)
I've never seen a case where the only way to follow policy was to repeat text unnecessarily. Anyway, between the time the above suggestion was made and now, the guideline has been modified to emphasize policy far more heavily; I suggest looking at the current version. Eubulides (talk) 23:35, 16 March 2010 (UTC)

Alt-text length

The above mentioned FA discussion has found some RNIB guidelines on alt text that seem to recommend much shorter (one sentence) text. Perhaps the discussion could be continued here rather than on WT:FAC which has no business discussing WP-wide guidelines.

I wonder if the RNIB guidelines are more aimed at trying to please the lowest-common-denominator and for websites where the images are merely cosmetic decoration (the image of a lady having difficulty reading a letter) or adverts (the DELL image), such as a shop or a government website. Perhaps someone can contact the RNIB or other appropriate bodies to consult with them how their members would like an encyclopeadia to handle alt-text for its images. The advice might differ depending on the use of the image, such as a painting, a photo, an illustration or a piece of design like an album cover. I suspect we should be more elaborate than a sentence or shorter ("a dog", "a person"). Due to the difficulty in finding images, our image tend to be more than merely decoration. Colin°Talk 20:26, 12 March 2010 (UTC)

Consulting with the RNIB sounds like a good idea. I agree with your analysis of that web page. A better page for the way Wikipedia typically uses images is Guidelines on alt texts in img elements, particularly its sections "When an image says more than a thousand words", "Making text and image really alternative", and "On the length of alt texts". Eubulides (talk) 21:00, 12 March 2010 (UTC)
I also think some consultation with experts would be helpful, for example with this chap, [1] who has a useful little online course about alt-text. I've done a bit of quick reading about the matter, and my initial impression is that the descriptions are to too detailed, and in a good number of cases something similar to the caption might actually be the best alt-text. The Korpela material above also emphasises brevity (50 characters if possible) for most occasions. I find these questions from W3C interesting [2]
"In deciding what text to include in the alternative, it is a often a good idea to consider the following questions:
  • Why is this non-text content here?
  • What information is it presenting?
  • What purpose does it fulfill?
  • If I could not use the non-text content, what words would I use to convey the same function and/or information?"
When thinking about function/information, in my view, a fair number of pictures included in WP articles that are tangential at best, and actually add precious little information. I think we may be overdoing the description in these cases, based on what I've read. But I think we really would benefit from some expert input related to the specific issues here. Too long appears to be as bad as too short. --Slp1 (talk) 22:18, 12 March 2010 (UTC)
I'm going to have another read of this guideline and the above linked external guidelines tonight if I can. I may then try emailing the RNIB to see if anyone would be interested in helping us here. Colin°Talk 13:37, 15 March 2010 (UTC)
I'm going to write to the guy I mentioned above, and see if we can get him involved. I really do think some expert voices will be helpful here.--Slp1 (talk) 14:29, 15 March 2010 (UTC)
I've emailed webaccess@rnib.org.uk. Colin°Talk 22:32, 15 March 2010 (UTC)

Context

I think our "Context" section needs to be expanded to include the length/detail changes for different circumstances. For example, the image of Tourette in Tourette syndrome is not key to the subject (an illness). The same image in his biographical article (Georges Gilles de la Tourette) is clearly of stronger encyclopeadic value. I don't think our guideline currently suggests that a description of Tourette's appearance could/should go in the body text of his biography. Perhaps editors should be encouraged to consider how weaker their body text becomes without the pictures and to consider whether alt-text is the most suitable place to make up the loss. The guideline currently suggests long descriptions go in the image summary. Since this is often a template covering multiple languages and other aspects like copyright, is this actually a friendly place for a screen reader to go, and would any blind editor consider clicking on the link for an image to learn more about an image they can't even see? Colin°Talk 13:37, 15 March 2010 (UTC)

I agree, based on what I have read to date, that the "value-added" of images needs to be considered. Your example is excellent, in that Tourette's appearance is scarcely relevant in the article about Tourette's syndrome. Another example is the article I am currently working on, Olivia Manning, where the pictures I have added are by (copyright issues) necessity fairly tangential to a reader's understanding of the topic.--Slp1 (talk) 14:29, 15 March 2010 (UTC)
Thanks for mentioning this. I added to Context the following sentence:
'Context also affects how much brevity is needed: for example, if an article about the man relied on this image to convey the man's appearance to the reader, the image's alt text might give several details about the man's features, whereas the same image in the article Blue might need only "A policeman in a dark blue uniform".'
Further improvements are welcome. Eubulides (talk) 18:48, 15 March 2010 (UTC)

This web page is a most interesting one, and emphasizes the role of context, very, very strongly. I suspect the issue of context needs to be moved right up to the top in our guidelines. Take a look at the interesting and quite relevant examples. Once again, it is noteworthy to see that the alt text suggestions are very short, and captions are considered good substitutes at times. I'm going to try and get in touch with these guys too. Can others try and get external reviews too? --Slp1 (talk) 22:19, 15 March 2010 (UTC)

I fully agree that probably context is everything and should be the #1 consideration. Captions often have to serve a naming role rather than a descriptive role, so might not be as amenable to being used as an alt-text substitute as with some websites. However, perhaps we need to cover what to include in the caption and what in the alt-text such that they complement each other.
We might also have to think imaginatively about how to handle the "longdesc" option should a lengthy description be desirable yet perhaps conflict with the rules for alt-text. The current text mentions the image's page summary text, but I don't think that is appropriate. Colin°Talk 22:32, 15 March 2010 (UTC)
Half answer, I find it interesting that in the examples about George Washington, where the text is obviously about George Washington, with a caption giving his name, an empty alttext is recommended. The naming role seems all that it recommended.--Slp1 (talk) 22:53, 15 March 2010 (UTC)
The examples on that page are all iconic, so they're not good examples for the guideline here. I agree that context is crucial (something that the page says but does not illustrate, alas), and it would be fine with me to move it up. The guideline already attempts to say what should be in caption vs alt text, in the Goal and Context sections (as well as in the lead), but no doubt this should be improved. I suggest also looking at this page's "Complex images" section. The first preference (put everything in adjacent text) is good for blind but bad for sighted readers, the second preference (longdesc="#footnote") would be good but MediaWiki doesn't support it, the third preference (put it on the file page) is what WP:ALT suggests now, the 4th preference (long alt text) is recommended-against by WP:ALT, and the 5th preference (longdesc=some other page) is also not supported by MediaWiki. I imagine that one could find circumstances in which each of the 5 choices would be reasonable ones. I wish MediaWiki supported longdesc. Eubulides (talk) 22:53, 15 March 2010 (UTC)
I'm not so sure about the influence of iconicity myself. The discussion regarding the Delaware article, and what the best choice would be best depending on context of the article is a great example to me. I found it most interesting in your link that disabled and nondisabled users had such different responses about what was desirable.[3] --Slp1 (talk) 23:08, 15 March 2010 (UTC)
Here's another interesting version of recommendations, complete with some examples. [4]--Slp1 (talk) 23:37, 15 March 2010 (UTC)

Guideline

A lengthy discussion at Wikipedia talk:Featured article candidates#Alt text in FA images has covered two main areas of disagreement:

  1. Does the guideline conflict with our policies, particularly WP:OR?
  2. Is the advice sufficiently wise and mature and in keeping with recommendations by external authorities?

Many participants at that discussion feel these issues mean it should not currently be enforced at FAC, though that in turn raises the question over whether it should currently be a guideline for any WP content.

Efforts are underway to bring in external advice to help resolve the second of these issues. There are aspects of WP's use of images that differ from most websites, which may use stock photos or many graphic-design elements.

I would much prefer we work together to fix and improve this guideline than demote it, even temporarily. Colin°Talk 21:17, 15 March 2010 (UTC)

I expect my input is unwelcome here, as I have made it very clear I believe that the present alt text guidelines are ill-conceived, and until they're aligned with best practice ought not to be imposed on FA candidates. Nevertheless I can see that are moves afoot to elicit external expertise and perhaps to address the restrictions imposed by the wikimedia software in the implementation of features like longdesc. I am all in favour of a helpful and well-thought alt text guideline, just as I'm convinced that the present one isn't it. --Malleus Fatuorum 23:27, 15 March 2010 (UTC)
Malleus your input is welcome. I have felt more informed having read your posts over at WT:FAC. You seemed to have a thing against long text encourages original research which others have not raised. Regards, SunCreator (talk) 01:40, 16 March 2010 (UTC)
Several others made the same point before I did, but they've been more sensible than me and kept their heads down so far. I on the other hand have nothing to lose by telling it like it is, as I have nothing anyway. --Malleus Fatuorum 02:06, 16 March 2010 (UTC)

I would outline my issues with WP:ALT here as well. As I see it the primary question that needs answered is this: Does WP:V apply to alt texts, and if so to what degree? The various sub issues are these:

  1. Is it a violation of WP:OR to put information in a alt text that is not made explicitly obvious within the image?
  2. Does WP:PRIMARY apply to alt texts when the information included in the alt text is not also included with information available in the same source as the image? (For example, an image scanned from a book that does not put the image in any context)
  3. If WP:PRIMARY does apply, then are self created images acceptable sources as they are not published elsewhere? (WP:PRIMARY says primary sources can only been used when they have been published by a reliably third party publisher)
  4. Is it a violation of WP:SYNTH when speculations are made in alt texts? (For example, guessing at someone's age based on their appearance)
  5. Does WP:ALT encourage violation of these policies with statements within WP:ALT? (For example, "The alt text should be concise and should emphasize the image's most important visual aspects: it should summarize the essence of the image rather than describing every detail." - is an editor qualified to decide what the most important aspects of Mona Lisa? Or does such an assessment require a source?)
  6. If reliable sources are being used to source the alt text, how should it be cited, if at all?

I think this summarizes the core policies at conflict with WP:ALT. It is my opinion that anything more than the most basic description of the image, without a reliable source, is a violation of WP:OR. Before anyone responds to my issues, please read over the policies I've linked too, they are each fairly brief and to the point. Think about what the ALT guideline is telling us to do, and ask yourself if it is in compliance with those policies. If those guidelines authorize WP:ALT in its current form, there should be a shred of text in those policies authorizing some aspect of the alt texts. I've read them all many times, the policy does not support alt text in its current guideline format. So we need to decide if WP:V has no application to alt texts. If it does apply, then we need to rework the ALT guidelines to be compatible with those policies. —Charles Edward (Talk | Contribs) 02:28, 16 March 2010 (UTC)

As an addendum, I would point out that WP:V says: "This is strictly applied to all material in the mainspace—articles, lists, and sections of articles—without exception," which would include alt texts. I think the current view towards alt text is backwards. We are looking at is as someone just describing an image to our readers. It is more than that, we are using a source (an image) to write a bit of text (the alt text) to convey information to a reader. When you term it like that, it is plainly obvious that the current alt texts being employed are a subtle violation (and in some cases gross) violation of policy. Either way you shake it, WP:ALT has to give or WP:V has to give - they are currently in conflict. WP:V says ALT texts must meet the same sourcing standards as everything else, while WP:ALT says they do not. —Charles Edward (Talk | Contribs) 02:39, 16 March 2010 (UTC)

  • Comment. The previous comments are incorrect about verifiability, as WP:ALT#Verifiability states that alt text has to meet the same WP:V standards that everything else has to meet. An editor who summarizes a cited textual source in Wikipedia should use the same verifiability standards as an editor who summarizes an image. In both cases, editorial judgment is required, but exercising editorial judgment is an ordinary part of Wikipedia editing. Eubulides (talk) 04:02, 16 March 2010 (UTC)
File:Ian Tomlinson speaking to police.jpg
Ian Tomlinson moments before he was struck
  • The three core content policies (NPOV, V, and NOR) apply to everything in an article, and ALT text is obviously problematic. The longer and more elaborate it is, the more it risks violating the policies, and it almost necessarily violates NOR because the description is always just the opinion of a Wikipedian. With simple images ("dog throws stick") that doesn't matter, but with complex or contentious images, it might. I've mentioned the image on the right as an example before, and it's one of a series of images I've had difficulty describing without violating the policies. Ought I to have described the slightly snarky look on Ian Tomlinson's face as he speaks to the police? Not to describe it is to leave out an important part of the story, given that one of the officers hit him moments later. But if I wanted to describe that look, what is it exactly? Snark, anger, frustration, tiredness? Toothache? We necessarily engage in OR when we add ALT text to images like this, by act or by omission. SlimVirgin TALK contribs 05:09, 16 March 2010 (UTC)
A reply to: "The previous comments are incorrect about verifiability, as WP:ALT#Verifiability states that alt text has to meet the same WP:V standards that everything else has to meet." - No, WP:ALT is saying every image must have an alt text and it must be fluffed up to convey the images essence or significance, and that is how it is being applied at WP:FAC. Under WP:PRIMARY, an unpublished source cannot be used to source anything, and any text cannot go beyond the limits of what its source says, or shows. So if my alt text is saying anything not explicitly made known in the image itself or a secondary source, (as is done in many of the example alt texts) you are violating WP:OR. I know that the disclaimer is there, but that does not change the what the rest of the guideline is advocating. Its like putting a disclaimer at the front of a manual on robbing banks saying all bank robberies must not break any laws. —Charles Edward (Talk | Contribs) 15:41, 16 March 2010 (UTC)
  • Saying that alt text "almost necessarily violates NOR" is like saying that any text any Wikipedia editor writes in an article "almost necessarily violates NOR". For example, the article text that appears next to the abovementioned image in Circa 7:15 pm: First alleged assault is problematic in the same way that the alt text is problematic. That article text says that Tomlinson "continued walking" and that "Police officers are reported to have followed him", phraseology that implies more doubt about whether police officers followed him than about whether he continued walking, and yet the cited source does not use wording conveying this extra doubt. Shouldn't the text be changed to say that Tomlinson was "reported to have continued walking"? Omitting "reported" leaves out an important detail; is that the right thing to do? This sort of editorial decision needs to be made with ordinary text all the time, just as it needs to be made with alt text all the time, and there's nothing special about alt text in this respect.
  • By the way, the existing alt text for that image is quite long: it's easy to make a case that it runs afoul of WP:ALT #Brevity. That's a different topic, though.
Eubulides (talk) 08:02, 16 March 2010 (UTC)
"Reported to have followed him" describes that T walked, and the police walked, and it was reported that they were not just walking but following him. Please stick to the ALT text issues, because there are no sources there at all. I'd appreciate it you would address the point about the expression on his face. SlimVirgin TALK contribs 08:18, 16 March 2010 (UTC)
There are just as many sources for the alt text in question (one image) as there are for the article text in question (one news article). In both cases, editorial judgment is clearly present in the choice of what text is present and what text is absent. I see no need for the alt text to describe the expression on T's face; the alt text is already way too long as it is. Eubulides (talk) 09:09, 16 March 2010 (UTC)
But in failing to describe the expression on his face, but in describing other things, I'm engaged in OR. I've decided all by myself that it's not relevant. In addition to being OR it's also arguably an NPOV violation. I don't think there's any way to write ALT text for that image without violating the content policies, unless we keep it very brief ("several police officers with dogs, and a man, looking at each other") and even then ... SlimVirgin TALK contribs 09:16, 16 March 2010 (UTC)
This argument appears to contradict itself. It's saying that it might be OK to use brief alt text like "several police officers with dogs, and a man, looking at each other", but that this brief alt text constitutes original research because it describes some things but not the expression on the man's face. The simplest and most natural way to escape this seeming contradiction is the obvious point that it's not original research to describe an image merely because one gives some but not all details of the image. Eubulides (talk) 10:16, 16 March 2010 (UTC)
I also question whether this is a guideline in any meaningful sense. The guideline tag was added by Beland in December 2005 [5] with no discussion. Although the page has existed since 2003 I had never seen an image with ALT text until I saw it had been mandated at FAC, and there seems to be no consensus for that either. If people aren't doing it then it falls at the "policies and guidelines must also be descriptive" hurdle. There's been no attempt to inform the editors who read the NPOV, NOR, and V policies to see whether they object. People who do raise objections have been attacked along the lines of "SlimVirgin doesn't want to help blind people!!" which makes people not want to speak out. It's worth stressing that no one on this page has engaged in comments like that, but it has happened elsewhere. All in all, very problematic. SlimVirgin TALK contribs 06:26, 16 March 2010 (UTC)
There is surely no need to inform every policy talk page whenever a guideline is written, nor should there be a need to cast doubt on the validity of a decision that was made long ago and never seriously questioned in the meantime. If there are valid NPOV/NOR/V concerns now, they can of course be brought up on the relevant talk pages. New featured articles have consistently had alt text for several months, so the guideline for alt text in featured articles has been descriptive in that sense. There is indeed a desire to make featured articles accessible to blind and visually impaired readers, and a pretty clear (though not universal) consensus that WP:ACCESSIBILITY is a plus for the encyclopedia. Eubulides (talk) 08:02, 16 March 2010 (UTC)
E, there was never a decision made to promote this to guideline. One person simply added the tag without any kind of discussion, and it wasn't questioned because it wasn't noticed. And yes, if we're going to say it's okay and even mandatory to add potential OR, V, and NPOV violations to FAs, the editors who watch those policies need to be informed, because as things stand this guideline contradicts the content policies. And the point about alt text being descriptive at FACs it a little unfair; people add it to FAs only because they feel they have to to get through the process. Whenever it's discussed it's clear there's a lot of disagreement about it being in the FA criteria. SlimVirgin TALK contribs 08:14, 16 March 2010 (UTC)
Lots of people have noticed that this was a guideline in the past five years, and it hasn't been seriously questioned as a guideline until now. Nobody is saying that it's OK or mandatory to violate policy in FAs. My comment about the alt text guideline being descriptive for FAs was intended as a corrective to the claim that "people aren't doing it". I agree that there is not universal consensus about alt text, just as there isn't universal consensus about a lot of things at FAs. Eubulides (talk) 09:09, 16 March 2010 (UTC)

I also had not heard of WP:ALT until an article I was nominating for FAC was reviewed. Based on my experience, WP:ALT seems to be less of a Wikipedia guideline than a hoop to jump through at FAC. I am also concerned about points that other people have raised here and at WT:FAC: WP:ALT seems to often encourage original research-type descriptions (for example, Wikipedia:ALT#Portraits, with guesstimate ages listed for the people in the images), and the guideline probably should be rethought before enforced anywhere, particularly at FAC. Firsfron of Ronchester 06:53, 16 March 2010 (UTC)

Featured articles have more hoops to jump through, true, but this one is one of the easier guidelines to satisfy. Those women's ages do seem to be a sore point, so I boldly removed them. Are there other specific examples that you'd like to draw our attention to, to help improve this guideline further? Eubulides (talk) 08:02, 16 March 2010 (UTC)
Using "garbed" instead of "dressed"; describing things as "Edwardian" or "Brutus style," when lots of people won't know what that means. Going into a level of detail that is subjective, and that very few people would find helpful and might find annoying e.g. deep-set eyes gazing solemnly, "serious and dignified," "haunted expression." SlimVirgin TALK contribs 08:10, 16 March 2010 (UTC)
"Garbed" is a perfectly reasonable word, but if it annoys you, please feel free to change it. The "Brutus style" is explained immediately as "cropped close but with a short fringe in front", and the "Edwardian style" is similarly explained nearby, so I don't see a problem there. More generally, there's nothing wrong with alt text using a few words that typical 12-year-olds won't know, so long as they get the general gist (this is true for article text as well as for alt text, of course). The "deep-set eyes gazing solemnly", "serious and dignified," and "haunted expression" phrases are useful for conveying the essence of the images, which is more important than trivial visual details. If these phrases can be improved upon without losing the images' gist, please feel free to do so. Eubulides (talk) 09:09, 16 March 2010 (UTC)
Garbed is a kind of silly word, and it's pointless to add Edwardian and Brutus if they have to be explained. And it's not just typical 12-year-olds who might have issues—huge numbers of our readers will have no idea what Edwardian refers to when it comes to Cameo brooches by the neck, or what a Brutus-style haircut is. Why not keep it very simple? No cliches, no leaps of interpretation, as little of the subjective as possible. If you want to call the image a primary source, it should only be used to make descriptive statements, never interpreted, per NOR. SlimVirgin TALK contribs 09:28, 16 March 2010 (UTC)
Again, if "garbed" annoys you, please feel free to change it. It is not pointless for an encyclopedia to explain things and to introduce new terms to people. That is what encyclopedias are for. It's fine to keep things simple and to avoid cliches in alt text, just as it's fine to do that in article text. It's also fine to sometimes introduce some details that are not simple, in alt text as in article text, if that helps conveys the gist of the image. Eubulides (talk) 10:16, 16 March 2010 (UTC)
I don't have an issue with "garbed", and don't see the problem with that particular word. I appreciate the changes you, Eubulides, made to the first two ALT text image templates. I see SV's point about NOR when it comes to phrases like "a haunted expression" (etc), but I do not share her view on that issue. Firsfron of Ronchester 23:55, 16 March 2010 (UTC)

NPOV

Sometimes in the interests of NPOV it's important to let images speak for themselves. The Tomlinson image above is an example of that. Indeed that's sometimes why we add images in the first place, because they tell stories that can't be told with words alone or can't be told with words at all. Alt text undermines that principle (a variant of "show don't tell), and removes from writers the choice about when to illustrate with text and when to do it with pictures. I'd support a guideline saying alt text is encouraged, and the FA criteria saying the same, but I think the decision should always remain with the writers. SlimVirgin TALK contribs 08:44, 16 March 2010 (UTC)

If images could speak for themselves, blind readers wouldn't need alt text! Alt text does not in any way take that choice away from writers: it doesn't affect the article at all, unless you're blind or otherwise visually impaired. The presence of alt text does not undermine the "show don't tell" principle one whit. All it does is help readers who would otherwise get nothing from the image. There is nothing harmful or POV about briefly describing the Tomlinson image to a blind reader. Eubulides (talk) 09:09, 16 March 2010 (UTC)
Or unless you're reading the article without images switched on, which could cover a lot of people. And I've explained above why I feel it's OR to include a description of his facial expression and also OR to exclude it. My point is that this kind of editorial decision should be left to the editors on the page who know the subject and who know what its NPOV pitfalls are. Nine times out of ten (dog catches stick), there won't be a problem. But in trying to force editors to add alt text to every image you're ignoring the small percentage of images that do introduce a problem. That's why one-size-fits-all is good to avoid whenever possible. SlimVirgin TALK contribs 09:21, 16 March 2010 (UTC)
My argument applies equally well to any set of readers who can't see the images: it doesn't matter whether they're blind or on a low-bandwidth connection. There's absolutely nothing wrong with including brief alt text for this image, and there is something wrong with not giving any information about the image to readers who can't see it. Eubulides (talk) 10:16, 16 March 2010 (UTC)
Any alt-text guideline should offer editors the choice to write "alt=", and indeed suggest that option in many cases. Even though we don't buy stock photos, many of our images are somewhat indicental to the article and may not warrant much or any alt text. And there will be rare cases like those mentioned above, where perhaps we dare not describe the image because it is just too sensitive (or editors disagree on how to describe the image). I'd rather that following the alt text guideline wasn't merely "encouraged", but that the guideline was sufficiently flexible itself. Colin°Talk 09:32, 16 March 2010 (UTC)
Yes, and the current alt-text guideline does suggest "alt=" in some cases, for purely decorative images. I've never run into a case where an image cannot be described at all, but if it's really a problem we could easily establish a convention for it ("alt=[Too controversial to describe]", say). Eubulides (talk) 10:16, 16 March 2010 (UTC)

Basic "alt" requirement

I've been working on the assumption that most folk accept it is a good thing for WP to be accessible and that following W3C standards on accessibility mean we should attempt to write useful "alt" text for images (which includes knowing when to say "alt="). As such, WP should have guidance on alt-text, whether that is a paragraph in WP:ACCESSIBILITY or a separate page such as this one. Is this something we can all agree on, or are there alternatives we should look at?

Let's assume there's enough to say about "alt" to warrant a separate page. What can we all agree on? Could we produce a few paragraphs, based on external guidelines and expert input if we get it, that most folk will be happy with. In other words, rather than let this guideline be dropped, we prune it hard back to something essential from which it could later slowly develop. I appreciate that this would be painful to those who have worked on the guideline, and that some good advice may get lost (hopefully temporarily) in the process. Is this a strategy worth pursuing?

Colin°Talk 09:32, 16 March 2010 (UTC)

The guideline is too large, and should be split into subpages like any too-large page. However, my impression from the discussion so far is that the core disagreements are about relatively small but crucial parts of the guideline, and that drastically pruning away large, less-essential parts of the guideline (or moving them to subpages) won't materially improve consensus. It would be better, I think, to fix these small but crucial parts. As the discussion illustrates, some of the examples in this guideline are key, and one way to help reach consensus might be to reword the alt text in these examples (some of this has been done already). Eubulides (talk) 10:16, 16 March 2010 (UTC)
I've just read the history section, which explains that, before 2008, the alt text was simply the caption. Given that the guideline tag was added in 2005, that means there has never been a consensus about having a guideline that says separate alt text must be added. Would it make more sense for this to be a proposed guideline? SlimVirgin TALK contribs 10:34, 16 March 2010 (UTC)
The guideline has always mandated alt text (Every time a Wikipedia article uses an image, the alternative text for that image should be specified. Only a software limitation prevented separate alt text in the case where the image had a visible caption. Otherwise, the user was required to write alt text and the guidance was different to that for a caption. I don't really see how lifting a software limitation requires the whole guideline be demoted to a proposed guideline.
I'm trying to find a way to keep the requirement for alt-text (which we've had for over four years, albeit mostly unenforced) and which I don't think many people really want to see dropped. The issue is the lengthy alt texts and some of the guidance on what to mention and what not to mention. If this gets dropped, WP has no requirement for alt-text and may not for quite some time. We will revert to churning out FACs with poor accessibility for blind and partially-sighted readers. Slim, I'm very glad you've come here as it is the only place progress can be made. The very worst option IMO, is the one currently beeing mooted at WT:FAC, that they collectively just choose to ignore the guideline. That would be a very sad step. Colin°Talk 13:50, 16 March 2010 (UTC)
Colin, I disagree with your last point. This shouldn't be mandated at FAC. If you care about alt text, forcing people to do it simply makes it unpopular. If you care about FA writers, you won't want to impose an extra chore on them. Either way, it hasn't been a good development. As for your other points, we've not had the requirement for four years if no one knew about it and it wasn't being enforced. Something's being tagged a guideline doesn't make it a guideline.
In addition, when Beland added the tag, the page said, "Unfortunately, the image syntax has no way of specifying alt text that is different from the caption. Therefore, if the image has a caption, please follow the advice in Wikipedia:Captions to choose an appropriate caption, and don't worry about the alt text ..." Now we're trying to force people to write additional words, and that's the thing that needs to gain consensus before this can be called a guideline. SlimVirgin TALK contribs 14:43, 16 March 2010 (UTC)
The "additional chore" claim has been responded-to elsewhere. I don't buy it. It just reflects the attitude that there's no enthusiasm for making the extra effort required to allow our best articles to be accessible to a wide audience. We mandate things at FAC that are a chore to get right, like image licencing templates, consistently formatted citations, reliable sources. It doesn't automatically make them generally unpopular. Why is such a contested guideline not got a talk page full of suggested/rejected changes? Why, if people are so angry about it, does it not have a history-log of page protection and 3RR warnings? Why is it criticised at length and in great detail from the comfort of WT:FAC, a forum where they aren't expected to actually suggest or make changes?
Anyway, I would appreciate an answer to the original question. "Should WP have guidance on alt-text, whether that is a paragraph in WP:ACCESSIBILITY or a separate page such as this one. Is this something we can all agree on, or are there alternatives we should look at?" Colin°Talk 16:30, 16 March 2010 (UTC)
WP should clearly do whatever is reasonable and practical to improve accessibility, whether that includea a separate page on alt text or not. The point that doesn't seem to be sinking in though is that the present "guidelines" are misguidelines, and need to be fixed before foisting this requirement on anyone. I have offered many examples of what I consider to be far better alt text than the examples on the present page, and I and others have demonstrated repeatedly with examples where alt text apparently written in accordance with the current practice fails several important WP guidelines. What more are others expected to do, when the only response from here is not to listen, and to continue to argue that doing anything is better than doing nothing? The discussion at WT:FAC is to do with whether, given the problems surrounding the imposition of this new alt text requirement at FAC, it should be removed until those who claim to know what they're doing sort it out. --Malleus Fatuorum 16:40, 16 March 2010 (UTC)
The previous comment appears to have been written without inspecting recent changes to this page, which include several changes along the lines that Malleus has suggested. Eubulides (talk) 18:57, 16 March 2010 (UTC)
I have seen the changes, but I don't think they cut nearly deep enough. --Malleus Fatuorum 19:02, 16 March 2010 (UTC)
What further changes do you suggest? I'm not asking for an exhaustive list, just some specific examples. Eubulides (talk) 20:27, 16 March 2010 (UTC)
Colin, I'd appreciate it if you and everyone else who wants alt text to be mandatory could drop the "you're just being lazy" or "you don't want to help blind people" arguments. Speaking for myself, alt text is practically the last straw for me when I'm writing an FAC and I have to force myself every time to struggle through it. This is for two reasons. It can be reasonably time-consuming, but more importantly I know that it's probably a waste of time, and it brings home to me powerfully that I'm not acting in my own interests, not being paid, not doing anything meaningful, and not enjoying myself. That is not a nice feeling. Frustration is a perfectly legitimate response and it's an appropriate topic to raise at the FAC talk page.
I'd have no problem with having sensible guidance on alt text that makes clear the content policies come first. I'd like to see it encouraged, not mandated; with exceptions for images that are too contentious to describe; and I think the suggestions should be less elaborate so they're easier to write and less likely to violate NOR and NPOV. SlimVirgin TALK contribs 16:58, 16 March 2010 (UTC)
I don't think you, Slim or Malleus, are in any way "lazy". There are few editors more hard-working at building this encyclopaedia. You've just described all the reasons (and they aren't invalid ones) why you have no enthusiasm for it. Why did that frustration not previously translate into coming here to this talk page and insisting it is too onerous, etc and suggesting the length should be cut down to one short sentence? Wouldn't that have been more useful than joining in the pile-on at WT:FAC?
Malleus, all your criticisms of the guideline mean it shouldn't be a guideline. Full stop. It would be trivially easy for you to torpedo this guideline as a guideline and that would be a better outcome for WP than just hiding the problem from FAC. There's a posse lined up waiting to "oppose" it. Its removal from FAC would then become a formality. I even started the process.
I'm sorry my frustration has led to accusations of people "not wanting to help". I see other guideline and policy pages being actively worked on by supporters and critics. With alt-text, I see a pile-on of criticisms but few suggestions of what to change other than that it be dropped till somebody else fixes it. I'm fairly certain I'm about to see WP take a step backwards rather than forwards (even if that forwards step was a bit wonky) wrt accessibility. I'm encouraged that some discussion has started here and changes are being made.
Sincere apologies to you both and the others. Colin°Talk 18:25, 16 March 2010 (UTC)
I have no desire to see WP "take a step backwards" in accessibility, but my feeling about the present guideline is similar to the reply the Irishman gave when asked for directions to Cork; "Well, if I was going to Cork, I wouldn't start from here." I have offered several examples of articles with what I consider to be good and useful (to a visually impaired reader) alt text, and hopefully with input from external experts a new guideline can be developed, one that's both helpful to the reader and not demanding interpretation by the author. Perhaps your idea of starting again with a cut down version, focusing on how to deal with the typical images encountered in an article is the way to go, and build from there. In the meantime editors making FAC nominations are getting frustrated at what appears to be the largely useless task of writing alt text that satisfies nobody and helps nobody. What should be our best work is ending up, well, not the best we can or should be doing, and not setting the example it ought to for other articles. --Malleus Fatuorum 18:40, 16 March 2010 (UTC)
Again, the previous comment seems to have been made without examining the recent changes to this guideline, which address issues raised in the comment. I don't see anything in Colin's comments that imputes laziness or coldheartedness to those who would like to improve the alt text guideline. Instead, Colin was saying, and I tend to agree, that most FA editors simply shrug and do the chore, hoping and assuming that it helps. SlimVirgin has recently rewritten the guideline to put the content policies first, which is fine. It's also fine to use somewhat terser wording in the examples, as SlimVirgin has been doing. Making alt text easier to write is a Good Thing. Eubulides (talk) 18:57, 16 March 2010 (UTC)
Not at all, as I said above. There needs to be a good clear out, not a bit of nibbling at the edges. And it's difficult to interpret a comment such as "The 'additional chore' claim has been responded-to elsewhere. I don't buy it. It just reflects the attitude that there's no enthusiasm for making the extra effort required to allow our best articles to be accessible to a wide audience" as anything other than a charge of laziness, or worse, a lack of compassion. --Malleus Fatuorum 19:04, 16 March 2010 (UTC)
I asked above for further clarification as to what needs to be cut. Lack of enthusiasm does not mean laziness. I am never enthusiastic about citing sources (what a pain citations are!) but that does not mean I am lazy or unscholarly. Alt text is similar. Eubulides (talk) 20:27, 16 March 2010 (UTC)
I'm coming late to this discussion. I don't claim expertise in this field but have had to make a range of online learning materials comply with UK disability discrimination legislation in RL. I would suggest looking at the resources available from TechDis and, in particular their pages on Adding value to image based resources, Adapting image-based materials etc and possibly contacting them for expert advice.— Rod talk 19:10, 16 March 2010 (UTC)
Thanks, I looked at those sources but found no details about how much alt text to add. Eubulides (talk) 20:27, 16 March 2010 (UTC)

How to default to the caption

What do we have to write to make the caption the alt text i.e. to omit alt text where the caption suffices, but to stop the screen reader from reading the file name? SlimVirgin TALK contribs 17:44, 16 March 2010 (UTC)

In ordinary (thumb) images the only way you can make the alt text be the caption is to repeat the text. This is inadvisable because a screen reader will read the text twice, contrary to the WP:ALT#Repetition guideline. In plain pictures omitted alt text does default to the "caption" (see Defaulting to caption), but the "caption" in a plain picture is not really a caption at all, but is title text (a different thing), so there's no repetition problem: screen readers normally omit title text. Eubulides (talk) 18:57, 16 March 2010 (UTC)
Also, an earlier version from January 2009 said that alt text was only necessary for images that added information to the article.
"Alternative text may not be necessary if the caption itself suffices to describe the image, or if the particular details of an image are irrelevant. A practical test is to ask whether anything would be lost if the image were blanked for all readers. If the image per se contributes nothing to the article, then no alt text need be provided."
Should we restore that? If we have that, plus the ability to default to the caption where the caption is descriptive enough, that would relieve a fair bit of the burden. SlimVirgin TALK contribs 17:55, 16 March 2010 (UTC)
That earlier text was designed for plain pictures, not thumbnails. Treatment of images (and their alt text) has evolved in the software since then, and in response the current guideline has moved the parts of that earlier text into different sections. The "particular details of an image are irrelevant" business is now described in Purely decorative images, so that it is in line with WCAG 2.0 guidelines. The "caption itself suffices to describe the image" part is now described in Defaulting to caption and in Placeholders. It may make sense to coalesce these somewhat-scattered sections. Eubulides (talk) 18:57, 16 March 2010 (UTC)
You can just provide a default alt text (alt="") as the screen reader will read the caption anyway. The other question you raise about images in which the details are "unnecessary" reopens yet another can of worms. Eubulides has frequently argued that such images are therefore by definition themselves unnecessary, and should be removed from the article. --Malleus Fatuorum 18:17, 16 March 2010 (UTC)
  • With an empty |alt=, the screen reader will read the file name, which is undesirable.
  • The previous comment misstates the thrust of my comments. I've consistently said that alt text should be omitted for purely decorative images, and a placeholder used for images described by caption or adjacent text; I have never argued that empty or placeholder alt text means that an image should be removed. There have been a couple of specific cases where I've said that an image itself was unnecessary, but that's a different matter: images are necessary or unnecessary for sighted readers independently of whether the images have alt text.
Eubulides (talk) 18:57, 16 March 2010 (UTC)
  • With an empty |alt=, the screen reader will read the file name, which is undesirable. A screen reader will read out the file name if no alt text is provided, but I said "default alt text", the purpose of which is to tell the screen reader to ignore the image. --Malleus Fatuorum 19:11, 16 March 2010 (UTC)
  • In HTML is no such thing as "default alt text" that would cause a screen reader to ignore an image. In MediaWiki, to tell a screen reader to ignore a plain image, use both empty |alt= and empty |link=. Merely using an empty |alt= does not suffice. There is no way to tell a screen reader to ignore a thumb image. Eubulides (talk) 20:27, 16 March 2010 (UTC)
  • This is not an HTML issue, this is a screen reader issue. Default alt text is specified either as alt="" or alt=" ", and is a signal to the screen reader to ignore the image. --Malleus Fatuorum 20:32, 16 March 2010 (UTC)
  • Yes, but only if the image is unlinked (that is, it is not wrapped inside an "A" element). If the image is linked, screen readers announce the link to the listener, reading the alt text aloud, or (if there's no alt text or the alt text is empty) reading the file name. All Wikipedia thumb images are linked, and plain pictures are linked unless an empty |link= is specified. That is why, to mark an image as purely decorative, one must use a plain picture with both |link= and |alt= being empty. Eubulides (talk) 20:50, 16 March 2010 (UTC)
Okay, so if we want to tell editors how to make the caption the alt text, in cases where it's descriptive enough (not talking about decorative images), we can tell them to write: "link=|alt=", is that right? SlimVirgin TALK contribs 06:19, 17 March 2010 (UTC)
No, that doesn't work. You must omit the |alt= entirely. And you must not specify an empty |link=. This issue is discussed in WP:ALT#When to specify, notably in WP:ALT#Defaulting to caption, WP:ALT#Links, and WP:ALT#Purely decorative images. Eubulides (talk) 06:42, 17 March 2010 (UTC)

Would this help to move the discussion forwards?

Would it help to move the discussion forwards if we focused on one or two of the images given as examples in the guideline? I've seen that SV has done some work today on Napoleon, so shall we start with that?

Here's what the alt text looked like yesterday:

Full length portrait of Napoleon in his forties, in high-ranking white and dark blue military dress uniform. He stands amid rich 18th-century furniture laden with papers, and gazes at the viewer. His hair is Brutus style, cropped close but with a short fringe in front, and his right hand is tucked in his waistcoat.

This is what it looks like today after some initial pruning:

Full length portrait of Napoleon gazing at the viewer, wearing a white and dark blue military dress uniform. There is furniture next to him laden with papers, and a clock behind him. His hair is cropped close with a short fringe, and his right hand is tucked inside his waistcoat.

I have serious problems with both versions, although slightly fewer with the second version than the first. Would it be likely to lead to a fruitful discussion if we were to analyse this alt text together, and I try to make clear exactly what my objections to it are, and what I'd propose instead? --Malleus Fatuorum 21:16, 16 March 2010 (UTC)

If so, let's start with a simple question; should the alt text be telling me what I (as a sighted reader) can see from the thumbnail, or what I might see if I clicked on it to get the full-size image? --Malleus Fatuorum 21:20, 16 March 2010 (UTC)

That alt text is in the context of the first image in Napoleon, so it should tell you what you see in that 200×329px version of the image. Eubulides (talk) 21:28, 16 March 2010 (UTC)
OK. Then I suppose it's OK that the alt text on the full-size version just says this: "This is a file from the Wikimedia Commons". So what significance does the clock have in the background? Why has it been chosen to ignore the colour of the carpet? I only see papers in the bureau to Napoleon's left, not on all of the furniture next to him. What will the colours "white and dark blue" mean to a non-sighted reader, even one who has gone blind in later years? Is it Napoleon's trousers that are blue, with a white jacket, vice versa, or some other combination of the two colours? Is his haircut really important? If so, why isn't the colour of his hair also important? Here's my suggestion, introducing the subject immediately in a single short sentence:

Napoleon leaning on a bureau laden with papers

--Malleus Fatuorum 21:45, 16 March 2010 (UTC)
I'm beginning to think that this last suggestion may well be just the ticket on most occasions. My view is more and more that article context, and the purpose of the image serves in an article should drive the alt-text. In an article about Napoleon, the portrait may serve a purpose to describe him (once in an article), in which case perhaps more info about his hair, clothes, posture (because it is iconic) may be informative. I don't think the clock or even the papers adds much really. But this same picture appears The Amazing Race 6 where is serves no discernible purpose at all. In this, the case, the alt text should be set to ignore image or at most be "Portrait of Napoleon". My view is increasingly, as stated in this guideline, [6] that context is everything, and that actually it is wrong give examples of appropriate alt-text without thinking about, and giving, the context too.
BTW I've had an email response from one expert in the field. He's reluctant to get involved directly, but has expressed openness to answer direct questions. I've responded with some short questions about desirable length, detail and whether context/purpose is where we should start. I'll let you know if/when I get a response.--Slp1 (talk) 22:14, 16 March 2010 (UTC)
BTW2. It appears from this that using colours is not verboten in alttext. It makes sense as there is a wide variety of people who use alttext and many have not been blind since birth.--Slp1 (talk) 22:26, 16 March 2010 (UTC)
I agree that "Napoleon leaning on a bureau laden with papers" is preferable. I only added the clock because I was removing something (18th century and the name of his haircut), and felt I ought to replace it with something, but I agree that a very succinct description is best. SlimVirgin TALK contribs 06:22, 17 March 2010 (UTC)
  • 'I suppose it's OK that the alt text on the full-size version just says this: "This is a file from the Wikimedia Commons"' I don't observe that behavior. When I visit commons:File:Jacques-Louis David 017.jpg the alt text is the supremely useless "File:Jacques-Louis David 017.jpg". This would not be good alt text in an article, but it's a different matter.
  • "Napoleon leaning on a bureau laden with papers" Brief alt text like that is appropriate for the first context mentioned in WP:ALT #Proper names for that image. However, Slp1 is right: it's not enough for the second context, the lead image of Napoleon. In the second context the image's main function is to show the reader what Napoleon looked like. The bright military uniform, the military haircut, and the hand tucked in the waistcoat are all important parts of his appearance.
  • What is very clear is that the context of the image determines what should be said about it. What is equally clear is that if Napoleon's physical description is germaine to the article then it should be described in the article, not in an alt text. --Malleus Fatuorum 22:42, 16 March 2010 (UTC)
  • Yes, context is crucial, which is why this page has a Context section (the section that the Napoleon example is taken from). Napoleon's physical description is important to the Napoleon article, but that article should not burden a sighted reader with the textual description of an image that can be rapidly viewed and understood. One of the nice properties of alt text is that it doesn't burden sighted readers; we should not abandon this important technical advantage. Eubulides (talk) 22:58, 16 March 2010 (UTC)
  • Yes, colors are OK in alt text. The current guideline mentions color in several examples, including the lead example.
  • The vast majority of blind people become blind after birth. WP:ALT#Goal already says this, though perhaps it could say it more clearly.
Eubulides (talk) 22:34, 16 March 2010 (UTC)
Even I wouldn't necessarily propose an absolute ban on colour – I've given the example of blue sky before – but in this particular case I don't see that it adds anything even for a once-sighted reader. Is Napoleon wearing a blue-and-white checked military uniform? Perhaps ask your (Slp1) expert to comment on the versions of the Napoleon alt text above?
There is a pretty active visually impaired editor whose input would be invaluable here as well. Has he been invited to comment? After all, he is the only representative of the target audience we have easy access to. --Malleus Fatuorum 22:38, 16 March 2010 (UTC)
I have been reluctant to ask him, because he intensely dislikes disputes. Also, it'd be better to use an example other than Napoleon because Napoleon is an iconic figure and is very much a special case. Eubulides (talk) 22:58, 16 March 2010 (UTC)
So pick another image then. I only picked Napoleon because it had been worked on today as a result of this ongoing "dispute". --Malleus Fatuorum 23:14, 16 March 2010 (UTC)
What sort of example do you want? An image of a person? Color or not? Regards, SunCreator (talk) 23:30, 16 March 2010 (UTC)
From an FA, or an FA that had the text in when passed or just any article? Regards, SunCreator (talk) 23:32, 16 March 2010 (UTC)
It's not me who's objecting to the example of Napoleon, therefore it's not for me to decide. --Malleus Fatuorum 23:33, 16 March 2010 (UTC)
I suggest the first image in WP:ALT. Might as well begin at the beginning. The context is the lead image for Flag. Eubulides (talk) 23:35, 16 March 2010 (UTC)
Okay, perhaps try File:Wilhelm_Steinitz2.jpg as an example. It's used in eight different articles(to test context). Two(1,2) of which are FA-class but never had alt text at time of review. Regards, SunCreator (talk) 23:42, 16 March 2010 (UTC)
Lest I be accused once again of looking at an "unofficial" version of the alt text, perhaps you'd be kind enough to copy it across here now, as I did for Napoleon above. --Malleus Fatuorum 23:44, 16 March 2010 (UTC)

Edit to Napoleon in the guideline

This edit reduced the alt text from 39 words to eight. The short text found favour among some above but Slp1's comment could be interpreted that it was too short for this usage (lead image in Napoleon) and Eubulides was very unhappy about it. We describe the desk ("bureau laden with papers") but not Napoleon. For this usage, that just doesn't seem right. This is an iconic pose. It is the lead image in an article on Napoleon, not antique furniture. Can we find a compromise? Colin°Talk 14:58, 18 March 2010 (UTC)

We're not describing Napoleon, we're describing the image; big difference. --Malleus Fatuorum 15:03, 18 March 2010 (UTC)
And it is an image of Napoleon, not a bureau. Colin°Talk 15:10, 18 March 2010 (UTC)
Plus it makes the sighted-person and well-read-person mistake that because you know what Napoleon looks like and typically wears/stands, you don't need to tell the reader about that bit of the image. Colin°Talk 15:17, 18 March 2010 (UTC)
No, it is an image of Napoleon standing next to a bureau. If Napoleon's appearance is important then it ought to be described alsewhere, properly attributed. --Malleus Fatuorum 15:51, 18 March 2010 (UTC)
I'm not asking for it to describe what Napoleon looks like (that should go in the article, I do get the point), merely that you've focussed on the bits you don't already know from your visual and reading experence. The fact that the bureau is "laden with papers" is of less importance in this usage and with this subject than what he was wearing (briefly) or how he is standing in this picture. I agree we've got to keep it short but describing the bureau is "missing the point". Nobody stuck that picture at the top of the article because it was a nice one of a bureau laden with papers and happened to have Napoleon standing next to it. Colin°Talk 16:02, 18 March 2010 (UTC)
That the bureau is "laden with papers" may well be more important than the fact that Napoleon's wearing a blue and white military uniform; perhaps it was intended to demonstrate how industrous Napoleon was, for instance. The question you need to be asking yourself is this: "What information would be missing from the article if the image was removed?" Arguably none. --Malleus Fatuorum 16:44, 18 March 2010 (UTC)
They might have, but I doubt it. The answer to your question is not none. The image is not a stock photo used to illustrate an article on famous men. The loss of this image in this article would be greater than we could possibly describe in text. It is a wonderful and clear portrait of the subject of the article that captures a great deal of the man. Jacques Louis David did not commission a picture of a bureau. The alt text here needs to be brief, I'm not arguing that, but we've focused on the wrong thing. Anyway, I'm starting to repeat so I'm not going to push this further. Colin°Talk 17:19, 18 March 2010 (UTC)
(update) I've just noticed SV has applied a compromise and it is good. I hadn't noticed it among all the many edits going and wouldn't have continued here if I had. Colin°Talk 17:42, 18 March 2010 (UTC)
I'm happy with the compromise on the Napoleon text. I originally put Malleus's version as that imo was much better then the long version previous given, but was happy when SV changed it shortly afterwards. Regards, SunCreator (talk) 19:07, 18 March 2010 (UTC)

Flag example

The image is used in Flag:

A flag flies from a flagpole against the sky. The flag is red, divided in four by a white cross, the vertical bar of which is shifted toward the flagpole.

Taken from the lead section of WP:ALT. Dabomb87 (talk) 23:51, 16 March 2010 (UTC)

Added section heading. Regards, SunCreator (talk) 23:53, 16 March 2010 (UTC)
The detailed colouration of the flag is clearly irrelevant in the context of the flag article; anyone who wants to know can follow a link to see the description. That the flag is flying "against" the sky is also kinda strange. As opposed to "with" the sky? And who "shifted" that vertical bar? Make them shift it back at once! The alt text should be:

Flag of Denmark

--Malleus Fatuorum 00:22, 17 March 2010 (UTC)
(ec) It would be interesting to offer several alternatives for the same picture so that the editor can make a choice of the length and detail. The caption should also be available to him. How about these as possible alternatives to offer?
  • A flag
  • A red and white flag.
  • A red flag divided into four by a white cross
  • A flag flying; it is red and divided into four by a white cross.
  • A flag flying from a flagpole; it is red and divided into four by a white cross.
  • Flag of Denmark (added at edit conflict)
Tweaks or additions welcomed. --Slp1 (talk) 00:30, 17 March 2010 (UTC)
Most of those are meaningless though. So it's a red and white flag? So is the flag of Japan, so what do I learn from that? Divided into four whats by a red cross? --Malleus Fatuorum 00:36, 17 March 2010 (UTC)
Perhaps they are meaningless. I certainly don't like some of them. The point is to find out what the person who actually uses alt-text thinks, and for that we need a wide variety of choices, some we consider good, some we consider bad, so that we can develop a clearer understanding of "why" something is good or bad from his perspective.--Slp1 (talk) 00:53, 17 March 2010 (UTC)
There are more then one 'Flag of Denmark'. See Flag of Denmark, would it be worth considering distinguishing which flag it is? I'm not sure of wording to be more specific. Could say 'National flag' but there is more then one National flag of Denmark also. Regards, SunCreator (talk) 01:10, 17 March 2010 (UTC)
Good point. So why not just copy the whole Flag of Denmark article into the alt text then? I'm getting frustrated at having to keep making the same point, and I won't be doing so for much longer, I'll just be switching my attention to having this ludicrous guideline demoted and removed from the FA criteria. Let me spell it out once again. What in that image informs me that Denmark has more than one flag? The alt text is supposed to be a succinct description of the image, not an essay on Danish national iconography. --Malleus Fatuorum 01:34, 17 March 2010 (UTC)
This isn't rocket science folks. I'm with MF here. What's wrong with "a rectangular flag cut into four with an offset cross". What relevance do the colours have for people who can't see colours, and the ones who can can see the picture. Duh! --Fred the Oyster (talk) 01:57, 17 March 2010 (UTC)
The vast majority of blind people were formerly sighted, and know very well what colors are. When color stands out in an image (as it does here), it's entirely appropriate to mention it; this would be true even if the audience consisted only of people blind from birth. Eubulides (talk) 06:04, 17 March 2010 (UTC)
Let's get back on track here. These two last sections started off so well and productively. This process with the flag isn't about getting our opinions about what alt-text should be. We already have way too many of those opinions. It is about getting an example ready which we can use to get the feedback of a blind WP editor. BTW, Fred, you'll see above that multiple posts provide information about how colours can be used and useful for users of alttext. It is appropriate that this be part of the question to him. --Slp1 (talk) 02:09, 17 March 2010 (UTC)
Having already gotten positive feedback from a blind WP editor on the subject of alt text that I've written, I would like to caution you all that the above examples are unlikely to work well as a test. For one thing, the blind editor can't see the image. For another, the tests should be done in context, not in isolation. Eubulides (talk) 06:04, 17 March 2010 (UTC)
The point being that in most cases colours do not need to be used. A perfect point is this particular flag. What difference does it matter if it's red and white? The point is that it's a flag, and specifying its colour in an example of how-to is tantamount to saying that it's okay to use colours in alt text, when in the vast majority of occasions it isn't appropriate. What also doesn't seem to being taken into account of in these long descriptions is how long it takes for the screen reader to read them. The poor sod is there listening to a 5 minute thesis on a bloody flag. That's ridiculous. --Fred the Oyster (talk) 02:19, 17 March 2010 (UTC)
Perhaps in most cases colors do not need to be heard, but this case is an exception. None of the examples shown above would take more than three seconds to listen to, with a decent screen reader. Eubulides (talk) 06:04, 17 March 2010 (UTC)
Three seconds of their life that they are never going to get back and in addition to whatever is in the article. In most articles images are eye candy and are ultimately unimportant, as such it's vaguely ridiculous for it to have an extended and detailed description. For example the flag, if it's important to have the colour of the flag then why isn't that in the prose? And if it is in the prose then why does it need to be in the alt text? --Fred the Oyster (talk) 11:35, 17 March 2010 (UTC)
Malleus, your idea certainly has merit. How would you go about writing an alt text for a portrait? Try this example from the infobox image of recently promoted FA Ram Narayan (the image itself is File:Ram Narayan May 2007.jpg:

An old man sitting in front of an ornamented partition wall speaks into a microphone and holds a bowed instrument.

That's the current alt text. Dabomb87 (talk) 03:08, 17 March 2010 (UTC)
Well, from that picture I can't tell it's a "bowed instrument", it could be some kind of guitar-like instrument, the ornamented partition is irrelevant, and I can't tell whether Ram is speaking or singing into the microphone. And why so coy about using the man's name?

Ram Narayan in old age

--Malleus Fatuorum 11:56, 17 March 2010 (UTC)

Because the Alt-text page states "Alt text typically should not name people or objects in an image." and gives the example the example "Four men in suits sit smiling around a glossy wood table. A photographer hovers in the background." That explains the image content but doesn't give the names or brand names. That he sings or that the instrument is guitar-like I don't understand. Hekerui (talk) 14:11, 17 March 2010 (UTC)

"Flag of Denmark" is bad alt text for that context, since it repeats the caption, violating WP:ALT#Repetition, and it assumes unlikely expertise from the reader, violating WP:ALT#Verifiability. (One cannot expect the typical reader of Flag to know offhand what the Danish flag looks like.) Let's try the example again, with more context from Flag:

A flag flies from a flagpole against the sky. The flag is red, divided in four by a white cross, the vertical bar of which is shifted toward the flagpole.
The oldest national flag design still in use is Denmark's 13th-century Dannebrog.
A flag is a piece of fabric, often flown from a pole or mast, generally used symbolically for signaling or identification. It is most commonly used to symbolize a country. The term flag is also used to refer to the graphic design employed by a flag, or to its depiction in another medium....

In this context, it's clearly an error to say "Danish flag" since the caption already says that. However, the caption does make a point that this is the oldest national flag design still in use, which makes the design more important: what sort of design does a really old design look like? That's a reasonable question that a blind reader could well want to know, and the alt text should help answer that question. For this reason, "a flag" and "a red and white flag" are also poor choices for alt text; the others choices listed above are better, particularly the ones that mention "flying" since (aside from color) that's the most-arresting part of this image and what distinguishes it from a stylized flag design. Flags flying (as opposed to flags pasted onto screens) are the original use of flags, and this image is focusing on history as much as on the present. Color should of course be mentioned for this image, since the flag's color stands out and is one of the first things one notices about it. Eubulides (talk) 06:04, 17 March 2010 (UTC)

So, given the context, it seems that the reader should be told 1) the colour; 2) the pattern and 3) that it's "flying". A concise alt text statement would be something like, "A flag flies from a pole; the flag features a white cross on a red background." Everything else is irrelevant—the sky, the angle of the pole; at most, you might want to disambiguate from a diagonal cross. What else does it need? Steve T • C 09:39, 17 March 2010 (UTC)
Or even simpler, with the keyword first, "Flag with a white cross on a red background flying from a pole". --Malleus Fatuorum 11:43, 17 March 2010 (UTC)

Source request

Could someone who has access to these sources tell us what they say exactly to support these two sentences?

Absent or low-quality alt text is one of the top causes of frustration for blind users of the web.[1] A 2008 study of blind access to Wikipedia listed lack of alt text as the first obstacle.[2]

SlimVirgin TALK contribs 08:41, 17 March 2010 (UTC)


The Lazar text says:

"The top causes of frustration reported were (a) page layout causing confusing screen reader feedback; (b) conflict between screen reader and application; (c) poorly designed/unlabeled forms; (d) no alt text for pictures; and (e) 3-way tie between misleading links, inaccessible PDF, and a screen reader crash.

(my emphasis). So it is one of the top causes of frustration. It later breaks down the alt text frustrations into

  • No alt text for pictures: 18
  • No alt text for pictures-required registration: 5
  • Nondescriptive alt text: 10

The middle one is later described as a worrisome development and refers to the user of Captcha style techniques. It represents only 15% of the alt text issues and is pretty irrelevant wrt to alt text on our articles. I made this edit to better match the text to the source. Colin°Talk 11:38, 17 March 2010 (UTC)

From Lazar "Many of the most frequently reported causes of frustration are actually relatively easy for Web designers to solve. For instance, in the top causes of frustration discussed in this article, it is relatively easy to improve the labeling of forms, improve the labeling of links, add appropriate alternative text for pictures, improve the clarityof page layout, and make PDF files accessible"; "There were a number of causes of frustration related to alt text. In some cases, alt text was missing; in other cases, alt text was present but nondescriptive (e.g., “picture here”)"
Buzzi: "In the following, main accessibility and usability problems observed for both [English/Italian WP] versions are reported. The first problem concerns links. JAWS provides the list of all links in the page. If graphical links do not have any alternative description, the screen reader extracts (from the <src> element of the <img> tag) and reads the directory/filename." --Slp1 (talk) 11:49, 17 March 2010 (UTC)
Thanks to both. Slp, in the text you cite what is meant by "graphical links"? SlimVirgin TALK contribs 11:55, 17 March 2010 (UTC)
Could you justify this edit where you restore the "especially where anti-spam measures are in place requiring readers to describe an image to register with a site". Where does the source say this is "especially" the issue concerning alt-text? The source has figures showing it is only 15% of alt text issues, the source body text mentions it only once in the middle and not in the abstract nor the conclusions. When it does mention it, it mentions it last out of all the issues. And the solution isn't to write alt text but to supply an audio version of the Captcha image, to verify a human is present. This is all irrelevant to our guideline and sends out the wrong impression about where the major frustrations lie. Colin°Talk 13:05, 17 March 2010 (UTC)
The source says the spam thing is a major cause of frustration, and I think does say especially or words to that effect. I will look again. SlimVirgin TALK contribs 14:27, 17 March 2010 (UTC)

I see that the Wikipedia source has been restored. "Missing alt text is an important obstacle for access by the blind to Wikipedia." Buzzi M, Leporini B (2008). "Is Wikipedia usable for the blind?". Proc 2008 W4A. Beijing. pp. 15–22. doi:10.1145/1368044.1368049. {{cite conference}}: Unknown parameter |booktitle= ignored (|book-title= suggested) (help) Could someone please say what that source says about alt text and Wikipedia, apart from the three sentences quoted above? SlimVirgin TALK contribs 17:02, 17 March 2010 (UTC)

Basically nothing, I am afraid. (You have mail.) Hans Adler 18:14, 17 March 2010 (UTC)

Problems with guideline status

I've posted this on the FAC talk page. I'm posting it here too because I think it raises doubts about whether the page in its current form ought to have guideline status. I think we either need to remove that tag, or restore the spirit of the pre-June 2009 version, which made clear that not all images need alt text, and where they do have it, it should be brief and to the point.

  1. The page was created in 2003 to say that some images needed alt text, but only where the image's meaning was important: "If an image is an important and unique part of what you are trying to say, provide alternate text..." At that time, image captions defaulted to be the alt text (I believe).
  2. In December 2005, the guideline tag was added by Beland without discussion.
  3. In October 2008, a new function was added (I believe) allowing us to write separate alt text with the "alt=" parameter. The guideline continued to give the same advice as before. In March 2009 it said: "Alternative text may not be necessary if the caption itself suffices to describe the image, or if the particular details of an image are irrelevant. A practical test is to ask whether anything would be lost if the image were blanked for all readers. If the image per se contributes nothing to the article, then no alt text need be provided." It added that original research had to be scrupulously avoided.
  4. On June 23, 2009, Eubulides added to the FA criteria that images need alt text, which implied that all non-decorative images needed it, though even the guideline at that point didn't say that. [7] E first raised the issue for discussion on June 17. Others may read the discussions differently, but I read them as saying there was no consensus to add it. When he was reverted, he restored it on July 1. See the discussions here:
    Wikipedia_talk:Featured_article_criteria/Archive_9#Alt_text_in_images
    Wikipedia_talk:Featured_article_candidates/archive39#Alt_text_in_images
    Wikipedia_talk:Featured_article_candidates/archive39#Alt_text_helps_the_visually_impaired
  5. Beginning on June 24, 2009, Eubulides made a series of edits over several months to the alt text page that changed its emphasis. It now required all non-decorative images to have alt text. The section saying alt text isn't necessary for images already described in the text was removed. Examples of elaborate and interpretive alt text were added (naming certain kinds of hairstyle, saying which century furniture belongs to, describing facial expressions as e.g. "haunted"), thereby increasing the risk of NOR and NPOV violations. The ability to default to the caption was also obscured—not removed, but added low on the page, not included in the lead, and made quite hard to understand for non-technical editors. These changes may not have been noticed because the page has low traffic. In May 2009, just before the changes began, it had only 500 hits and the talk page 53. Even with the current extra attention it's only on 62 watchlists.
  6. Eubulides started requiring alt text at FACs and FARs. Editors added it when asked to because it's easier to do it than to argue, especially when you're under FAC or FAR pressure.

In summary, there seems (to me) to be no consensus to require editors to add separate alt text to all non-decorative images, mandate it at FAC, or recommend any kind of elaborate alt text where we go into detail about facial expressions and hairstyles.

Sorry, E. I had intended not to raise this again because I respect your work and I know you feel strongly about this, but given that others are raising it, it isn't fair to keep it going as though most people are on board with it. SlimVirgin TALK contribs 10:18, 17 March 2010 (UTC)

What to write for the caption to be the alt text

Can someone tell me what people need to write if they want the caption to be the alt text? Looking back over the talk pages, I see conflicting advice: add no alt text at all; add "alt=see caption," or "alt=|link=". SlimVirgin TALK contribs 12:41, 17 March 2010 (UTC)

There's no point in having the caption as the alt text, because a screen reader would read out the caption anyway. The problem as I see it is that there's no way to specify a default (i.e., empty) alt text for a thumbnail image, which is the ones we're almost always going to be dealing with in articles. --Malleus Fatuorum 12:59, 17 March 2010 (UTC)
I tested those three options. Only the second gives any alt text of 'see caption'. No really ideal. Regards, SunCreator (talk) 13:12, 17 March 2010 (UTC)
Or alt=""? Malleus, what I mean is if the caption is appropriate for alt text, what do we write to stop the screen reader reading the caption twice. SlimVirgin TALK contribs 13:35, 17 March 2010 (UTC)
Alt="" doesn't help with thumbnail images, as Eubulides pointed out above. The best we can do, it seems is to have an alt text that says "See caption".At least until the software is amended to deal with the expansion of thumbnails differently anyway. --Malleus Fatuorum 13:45, 17 March 2010 (UTC)
Nothing positive to report from testing. See User:SunCreator/Test and results Regards, SunCreator (talk) 13:56, 17 March 2010 (UTC)
Thank you! So is this wrong? "Where no alt text is supplied for whatever reason (for example, because the image is purely decorative), write alt="" or alt=|link=. Without that, screen readers will read out the file name in lieu of a description." SlimVirgin TALK contribs 14:02, 17 March 2010 (UTC)
Yes, it's wrong. This was discussed in #How to default to the caption above. Simply writing "|alt=" does not work, and "|alt=|link=" does not work for thumbnails. Eubulides (talk) 17:54, 17 March 2010 (UTC)
Thanks. Can you tell me what a person ought to write when they want the caption to be the alt text too i.e. where the caption is sufficiently descriptive, but writing it twice (once as caption, once as alt text) would be pointless, and we don't want the screenreader to read it twice. Please don't answer with, "But we mustn't do that." Please just tell me what a person would write if we were doing it. :) SlimVirgin TALK contribs 18:44, 17 March 2010 (UTC)
For a thumb image, the common case, you write "|alt=See caption.", as discussed in WP:ALT#Placeholders. For a plain picture you write the caption with no alt text, as discussed in WP:ALT#Defaulting to caption. Eubulides (talk) 22:21, 17 March 2010 (UTC)

Succinct in seven words

Matt Cutts of Google discusses the importance of alt Tags on YouTube, from official Google channel

"Really you don't need a ton of words because seven is enough to describe a scene pretty well, right? So, if you have twenty or twenty-five that's even getting a little bit out there."

— Author, Matt Cutts, Google

Discussion in the above section is moving in the direction of shorter succinct text, we have in the nutshell text "It should succinctly summarize the image's appearance". Google agree. So do I, here we have advice from Google with something quantifiable. Seven words is good, twenty or twenty-five and you are going off track. Regards, SunCreator (talk) 13:34, 17 March 2010 (UTC)

We have to remember that a Wikipedia article is not a web site. Imagine a style guide aimed at a monthly magazine being adopted for Britannica. It doesn't work. I reckon many web style guides tell their readers to ensure each web page takes no more than say 30 seconds to read, as the average user has no attention span. Would we apply that sort of logic to an FA? More like 30 minutes to read. Look at the RNIB home page. I see logos, some graphic art and a bunch of stock photos that merely decorate. That is a web site. This is an encyclopeadia. Colin°Talk 13:55, 17 March 2010 (UTC)
You make an improper distinction. It's not the images that are encyclopedic, it's the accompanying text. Very few images in wikipedia articles are strictly necessary to understand the text, but they add a little colour for those who can or choose to see them. --Malleus Fatuorum 14:04, 17 March 2010 (UTC)
I'm reading around about this, Colin, and every source I've found so far is recommending very brief alt text, including for significant images. SlimVirgin TALK contribs 14:05, 17 March 2010 (UTC)
I have avoided adding my two cents but the thing with succinct alt text is that its fairly easy to do in some cases but in others not so much. For example, I could describe fairly easily the ribbon for the Medal of Honor as "A light blue ribbon with five white, five pointed stars". To meaningfully descibe the the ribbon for the Silver Star is a bit harder, but it could be "A multicolored military ribbon. From left to right the color pattern is; thin blue stripe, thin white stripe, thick blue stripe, thick white stripe, thick red stripe, thick white stripe, thick blue stripe, thin white stripe, thin blue stripe." which is obviously very long. But otherwise how would a meaningful description sound? The other problem is that for someone how has been blind all their life what do the colors mean? It could be argued they mean nothing, however I have a couple friends who cannot see and they have informed me that although they cannot "see" the color, it does give them a sense of what they are "looking" at by describing that the colors change and how big the lines are. My point here is that we are making huge issues out of what meaningful alt text is but meaningful is going to be different depending on the disability and the person. --Kumioko (talk) 14:13, 17 March 2010 (UTC)
The thing is, who would want to listen to (and basically be held hostage by): "From left to right the color pattern is thin blue stripe, thin white stripe, thick blue stripe, thick white stripe, thick red stripe, thick white stripe, thick blue stripe, thin white stripe, thin blue stripe." Listening to text is not like reading it, where your eyes can scan and move away. If something's being read out to you, you're stuck with it unless you intervene to stop it, so we need to make sure we don't bore the reader stiff. SlimVirgin TALK contribs 14:18, 17 March 2010 (UTC)
I don't accept that few of our images have encyclopaedic value, but there are quite a number of usages where they don't and we should address that. I don't think the distinction is improper and don't see how one can fully "read around this" problem. Adopting mantra like "seven is enough", which is a web-site guideline, is a mistake. Note: I'm not fighting against the desire for brevity here. But our readers are patient enough to spend many minutes on an article, and they are here to find out information, not book a flight. Advice that applies to the graphic "Book now" button on the Easyjet site hardly applies here. I agree the above white strip example is ridiculous, but has anyone considered how long it takes to listen to an FA? That's a serious commitment. Colin°Talk 14:24, 17 March 2010 (UTC)

My only point is what Colin is saying in that some images are harder to describe than others. --Kumioko (talk) 14:27, 17 March 2010 (UTC)


Having finally managed to access this YouTube video, I'm just extremely puzzled that it was mentioned here. This guy is teaching you how to use alt-text for Google Image search optimisation. This guideline isn't even on the same planet as us. Let's keep things relevant. Colin°Talk 17:32, 17 March 2010 (UTC)

There is no way to ignore Googles influence even if it doesn't fit with your views. Matt Cutts is the authority behind Google's Webmaster guidelines which specifically includes Alt text and images. Regards, SunCreator (talk) 18:02, 17 March 2010 (UTC)
And those guidelines go no further than repeating the HTML requirement for alt text. This chap is not an accessibility expert, but he might be able to find where you left your car keys. Are you saying this guideline should have a section on search engine optimisation? Colin°Talk 18:23, 17 March 2010 (UTC)
I think the point is that the sources seem to be recommending brief alt text, and there's no reason to ignore them. I'm confused about your position here, Colin. You seemed earlier to agree that we needed to find advice for this page around which we could form consensus so that it could retain its guideline status. And it's clear that, insofar as there's consensus for any alt text to be mandatory, it would have to be brief alt text. Yet whenever that's mentioned you argue against it, but without supplying a source or any argument in favour of longer alt text. So we're just going round in circles. SlimVirgin TALK contribs 18:41, 17 March 2010 (UTC)
Did someone find a source on writing alt text for an online illustrated encyclopaedia, or even an information portal? All I can see are helpful sources (and one deeply unhelpful source -- above) but not authoratative ones because they aren't 100% relevant. There's not much we can do about that. I wish I could read some of the papers the IIT team (see below) have written. I think we should contact them. In the mean time, we can only do our best to combine what we've managed to learn from outside with own consensus and perceived needs. I'm comfortable with the consensus tending towards brevity but want us to keep our minds open to the possibility there are cases where a few sentences might be appropriate (the "flexibility" Hans mentions below). Colin°Talk 19:26, 17 March 2010 (UTC)
As Google's mission statement reads Google's mission is to organize the world's information and make it universally accessible and useful, with accessibility text here. I think at least you have to consider what they say seriously. Regards, SunCreator (talk) 19:57, 17 March 2010 (UTC)
If you can find some Google guidelines on writing alt text for people rather than robots, and written by an accessibility expert rather than a SEO expert, I might consider what they say seriously. Colin°Talk 22:41, 17 March 2010 (UTC)
I think the time has come then to demote WP:ALT to an essay and to remove the requirement for alt text from the FA criteria, as this discussion is going nowhere. --Malleus Fatuorum 22:46, 17 March 2010 (UTC)
Read accessibility text. It's all about people "We(Google) want to make information available to everyone, and that includes people with disabilities, such as blindness, visual impairment, color deficiency, deafness, hearing loss and limited dexterity" Regards, SunCreator (talk) 22:53, 17 March 2010 (UTC)
You do what you think is best Malleus, but don't use this particular section/discussion as an excuse. This is a distraction and doesn't reflect the good work and good suggestions being made elsewhere. I'm not going to repond further on this topic (Google's wonderfulness). Colin°Talk 09:03, 18 March 2010 (UTC)

The impression I got from scanning through the Buzzi & Leporini paper is that one of the main screenreader usability problems with Wikipedia is that it is confusing. E.g., it appears that screenreader users tend to cycle through the links rather than encounter them in their surrounding context. One of these links on the main page is labelled "more..." with nothing other than the previous text (often unknown to the screenreader user) indicating that it leads to the day's featured article. Apparently there are no less than 300 links on the English mainpage. (This seems to make overlinking of articles an accessibility problem; it might be worthwhile to remember this.) Under these circumstances it seems important not to increase the clutter for screenreader users with irrelevant alt texts. In my opinion alt text should be flexible: If a picture is the key to understanding an article, we may need several sentences (if possible technically and NPOV-wise). But in some cases a one- or two-word description may be preferable so it doesn't distract more than absolutely necessary. And often about seven words may be a good number. Perhaps what we need is a team of specialists who try to get this just right. Once they have developed a way of writing alt texts that works, they may be able to codify it. But I don't think it's very healthy to do that now, when we are still all a bit clueless. Hans Adler 18:49, 17 March 2010 (UTC)

And following on from that it's not appropriate to foist it half-baked onto FACs. I think though that the case for alt text always being short, perhaps a maximum of 20 words or so is irrefutable. Lengthy alt text is not an appropriate way to describe images that for whatever reason it's felt need to be dealt with in more detail; some kind of implementaton of longdesc is probably needed for them instead. To reiterate SV's point above, we need to be thinking of the poor reader who, every time an image with a long alt text appears on a page is going to be bored to death by hearing the same old lengthy speech again and again. --Malleus Fatuorum 19:24, 17 March 2010 (UTC)
Would it make sense to add to the page that seven words is a good rule of thumb, and that over 20 words will rarely be necessary (or words to that effect)? If it's presented as a rule of thumb it means people can ignore it if they feel they need to. As Hans says, if a team of specialists ever develops, the page can be tweaked according to what they find works best. SlimVirgin TALK contribs 19:33, 17 March 2010 (UTC)
Obviously I'd agree to that. It would at least serve as a sanity check against the streams of consciousness currently masquerading as useful alt text. --Malleus Fatuorum 19:37, 17 March 2010 (UTC)
These values are appropriate for functional images and advertisements (the common usage described in guidelines), but they are way too short for encyclopedic images, and do not reflect common practice on our best articles. Opinions on alt text styles differ, both among Wikipedia editors and among reliable outside sources. It's reasonable to suggest a range and a bound, but these should be adequate to cover commonly used styles: the guideline should not force editors to use a style on the very terse end of the spectrum. Eubulides (talk) 22:33, 17 March 2010 (UTC)
I think your comments display a deep misunderstanding of the nature and purpose of alt text, and further discussion seems unlikely to be productive. There is no such beast as an "encyclopedic image", so I wish you well with your soon to be demoted guideline, and hope that sanity will one day prevail. --Malleus Fatuorum 22:51, 17 March 2010 (UTC)
I'm going to add something about the 7-20 word rule of thumb, because we need to offer some idea of limits. SlimVirgin TALK contribs 09:06, 18 March 2010 (UTC)
I've added: "Its length will depend on the context, but it should be as succinct as possible so that readers are not burdened by unnecessary words—a good rule of thumb is that seven words is often enough, while more detailed images may require 20 or more." This offers guidance without pinning anyone down, which I hope is a good compromise. SlimVirgin TALK contribs 09:35, 18 March 2010 (UTC)
I'm happy with this. However, both context (mentioned at the start) and image detail (mentioned at the end) govern the length. How about moving both length-affecting variables to the front with: "depend on the context and complexity of the image, but..." and then "while some images and usages may require 20 or more". Colin°Talk 10:14, 18 March 2010 (UTC)
It may have changed since you wrote that, so maybe you could take another look and see whether it's okay for you. It's currently in the lead and the length and style section. SlimVirgin TALK contribs 12:59, 18 March 2010 (UTC)
  • I wasn't going to comment here, but I was reading the discussion above and I had to pipe up to object to Malleus' comment above: "There is no such beast as an "encyclopedic image"". That is clearly wrong. There are plenty of images that are crucial to help the reader understand what the article is describing (and I don't just mean diagrams), and such pictures are not eye candy. If anyone wants examples, please ask. Carcharoth (talk) 22:42, 18 March 2010 (UTC)

Example of problem

The problem with the current version, E, is that it doesn't really offer any guidance. The advice is inconsistent; the examples don't tally with the advice; and real examples of alt text written by you don't seem to fit the guideline. Here's an example I discussed with you last year, my first attempt at writing alt text. This scene is described by Wikipedia (in Clare College, Cambridge, which is the building on the left), as "one of the most celebrated architectural vistas in England."

1. My first attempt: King's College Chapel and the Gibbs building photographed from The Backs, looking east across the River Cam. Four people are punting on the river, the punter standing at the back of the boat on the deck or till, which is the style in Cambridge.

The caption concerned King's College, which is why I focused on that aspect of the image in the alt text, which I'd know not to do now. You said it was too specific, so:

2. My second attempt: King's College Chapel and the Gibbs building photographed from the River Cam. Four people are punting on the river, the punter standing at the back of the boat on the deck or till.

3. My third: "Several old buildings next to a river, including a chapel that dates to the 15th century. Four people are punting on the river, the punter standing at the back of the boat holding a long pole, while the others are seated."

4. Eubulides fixes it: Chapel in late Gothic style with a large window between two spires about eight stories tall, behind water and a green. Four people are punting on the water; the punter stands at the back of the boat and holds a long pole. On either side of the chapel are relatively nondescript three-story buildings.

I can't see why the guideline would disallow naming King's College Chapel, given that it's one of the most photographed buildings in the world, but allows you to mention late Gothic style, which will mean nothing to a lot of people, and to call Clare College, one of the most beautiful buildings in England, a "relatively nondescript three-story building." The guideline needs to set out some clear principles so that editors aren't floundering like this, because the uncertainty is very time-consuming when you're trying to figure out what's recommended. SlimVirgin TALK contribs 09:06, 18 March 2010 (UTC)

Actually, what is needed here is a better picture. Sometimes when you write alt text, you realise what is wrong with a picture in terms of framing and composition. The point about naming the building is that you shouldn't need to do that because it should be named in the caption. The problem is that the connection between what is said in the caption and in the alt text isn't always clear. Anyway, here is what I would write:
For a sighted person:

View of the west end of King's College Chapel (just right of centre), seen from across the River Cam. Between the river and the Chapel is the Back Lawn of King's College. The white building at right is the Gibbs Building, also part of King's College. At left is the south side of Old Court, part of Clare College. Just visible in the background, between Old Court and King's College Chapel, is part of the buildings of Trinity Hall. At lower left are people in a punt on the river, using a pole and paddle.

For a partially or non-sighted person, I would say:

View over a river of college lawns and buildings in Cambridge on a late summer's day. There are three college buildings, all in the middle distance occupying the middle third of the picture and forming the rear and left sides of a rectangular grassed area of lawn by a river. At left is a three-storey block in beige stone with the visible part of the building consisting of three rows of 17 windows. At right is a white stone and brick building with three rows of 3 windows on the visible part. Between these two buildings is the main subject of the picture, the river-facing facade of a tall chapel built in grey/beige stone: two tall towers, over six storeys high, framing ornate architecture, a very large window (several storeys high), and at ground level a large set of wooden doors. The chapel and the white building are angled to face the lawn (parallel to the river), while the building at left is angled 90 degrees to form the left side of the lawn. Other buildings in the far distance, are visible in the corner area between the chapel and the building at left. The foreground third of the picture is the lawn and river. The lawn is a large empty area of cropped grass, here seen from a very oblique perspective. Between the lawn and the river is a path running parallel to the river at the top of a short sloping grassed bank that leads down to the water. From this viewpoint, the path and riverbank are slightly angled from lower right to middle left. The visible area of the river, which is calm and tranquil, occupies the lower left area of the picture. Four people wearing summer clothes are in a punt on the river, moving from right to left. Two are sitting in the middle area of the punt, while one person at the rear of the punt is standing on the rear deck and using the punt pole, while the person at the front of the punt is using a paddle. The scene is bathed in bright sunlight and a clear blue sky forms the upper third of the picture. A tree, partially seen at far right, completes the scene, casting a shadow on the bottom right area of the river bank.

The punt is from Trinity College, but that isn't really needed for the caption or alt text. I've gone way overboard on the alt text (what I've done is more a complete description for a blind person, from which is is almost possible to draw the picture in your mind's eye), but if you compare the caption and the alt text, you will see how they are meant to work together. When writing one (caption or alt text), you should consider the other - they aren't really separate things. Compare the caption and the alt text point-by-point and you will see what I mean. And for some reason, people often forget to mention the weather and lighting conditions for pictures taken outside. I think adding that helps when describing a picture. And I recant what I said at the start of this post - it is not such a bad picture after all! Carcharoth (talk) 00:24, 19 March 2010 (UTC)

Best practice

There are three approaches we can use to formulate this guideline:

  1. Look at our own content and work out among ourselves the best approach.
  2. Read external guidelines (but see caution note above).
  3. Look at best practice.

Could we build a list of web sites that have a similar goal to us, and have adopted alt text. Here's one. Colin°Talk 15:04, 17 March 2010 (UTC)

This alt text for the image in the "Sports injuries" page is as follows:

Photograph of a teenage boy with a cast on his leg

"Photograph of" seems superfluous, but otherwise that seems to follow what WP's guidelines (and the other sources cited) on alt text, in that it briefly describes (but not interprets) the important aspects of the image. Dabomb87 (talk) 15:27, 17 March 2010 (UTC)

All three of the above and combine into our best approach. When considering any external views we have to consider Neutral point of view and No original research as that is not likely to be relevant outside Wikipedia. Regards, SunCreator (talk) 18:07, 17 March 2010 (UTC)

References and lead

E, it's a minor point but references don't have to be in the body of the article to be in the lead. SlimVirgin TALK contribs 15:49, 17 March 2010 (UTC)

Yes, they don't have to be, but the lead is supposed to summarize the body, which means that any claim that is in the lead should be repeated at more length in the body, and it's weird (and a sign of a weakness in a page) if different citations are used to support the same claim made in two different places. Eubulides (talk) 17:51, 17 March 2010 (UTC)
That's just not true. LEAD doesn't say it, and it's not best practice that I'm aware of. I've never heard the thing about refs before and I was using the same ref, not a different one, so I think we're talking at cross purposes. SlimVirgin TALK contribs 17:55, 17 March 2010 (UTC)

Duplicate citations

While we're on the subject of references, this page currently cites exactly the same source in "Notes" and in "References". I attempted to remove one duplicate citation, as there's no need to mention something in References if it's also in Notes, but was reverted, so I didn't try to fix the other one. The page shouldn't have unnecessary duplication like that. Let's use a single "References" section that combines the existing "Notes" and "References" section (since there's just a single function here) and that does not duplicate citations. Eubulides (talk) 17:51, 17 March 2010 (UTC)

My plan was to use short refs in the text and longer ones in the References section, but I was waiting until you'd finished editing. I'd appreciate if you wouldn't add templates to well-formed refs by the way, or leave out first names. If we know the full name there's no reason not to use it. SlimVirgin TALK contribs 17:54, 17 March 2010 (UTC)

Experts

Has anyone tried contacting the authors of "Is Wikipedia usable for the blind?": Maria Claudia Buzzi and Barbara Leporini? The team at IIT have several Wikipedia-related publications. BTW, I haven't yet heard back from the RNIB. Colin°Talk 18:03, 17 March 2010 (UTC)

Given the lack of reply, I'll assume no. I shall send them an email tonight (UK time). Colin°Talk 09:06, 18 March 2010 (UTC)

Source request again

I'm having problems with this claim: "... missing alt text is an important obstacle for Wikipedia access by the blind." [8] Sourced to Buzzi M, Leporini B (2008). "Is Wikipedia usable for the blind?". Proc 2008 W4A. Beijing. pp. 15–22. doi:10.1145/1368044.1368049. {{cite conference}}: Unknown parameter |booktitle= ignored (|book-title= suggested) (help)

A previous version of the sentence said: "A 2008 study of blind access to Wikipedia listed lack of alt text as the first obstacle." [9]

The only thing the source seems to say about alt text is: "In the following, main accessibility and usability problems observed for both [English/Italian WP] versions are reported. The first problem concerns links. JAWS provides the list of all links in the page. If graphical links do not have any alternative description, the screen reader extracts (from the <src> element of the <img> tag) and reads the directory/filename."

If there's anything else about alt text in that article, please let me know. We should summarize the source accurately or not mention it at all. SlimVirgin TALK contribs 08:39, 18 March 2010 (UTC)

The part about alt text follows later in the section containing that quote. That section says they looked at the main pages for Wikipedia, and found several problems, which they give in a bulleted list. The problems were:
  • Problems with links and alt text. (This is the bullet partly quoted above; it has the following two sub-bullets:)
    • Purely decorative images that are not properly marked with "|link=|alt=".
    • Images that lack alt text or have empty alt text.
  • Links with bad names (it's not specified whether these names came from alt text).
  • Too many links.
  • Unlabeled search edit fields.
  • Poorly marked section headings and levels.
  • Lists of links that are not marked as lists.
  • Graphical layout based on tables.
The alt text bullets take up most of the first column of page 17 (not counting the figures). At the end of the bullets (p. 18) the authors say that this list reflects research that was "aimed at highlight the main problems of interaction via screen reader." This is the basis for the claim "missing alt text is an important obstacle for Wikipedia access by the blind" you mention above. The claim accurately summarizes the source's claims about alt text under the constraints of brevity, but of course other wordings could be used as well. I see that you've removed the source again, which I find puzzling: the paper "Is Wikipedia usable for the blind?" is by far the reliable source that we have that is most directly relevant to the question whether alt text helps make Wikipedia more accessible to the blind. Eubulides (talk) 16:12, 18 March 2010 (UTC)
Agreed, but we need to summarize it accurately, and not make it say more than it does. SlimVirgin TALK contribs 16:19, 18 March 2010 (UTC)
If there's something inaccurate about the summary, then let's fix the inaccuracy, not remove the summary entirely. Eubulides (talk) 16:25, 18 March 2010 (UTC)
Could we deal with the issues below first, under Remaining issues? The section about text and unusual characters seems contradictory. SlimVirgin TALK contribs 16:30, 18 March 2010 (UTC)

The only thing I can find about alt text is on p. 17:

In the following, main accessibility and usability problems observed for both [English/Italian WP] versions are reported. The first problem concerns links. JAWS provides the list of all links in the page. If graphical links do not have any alternative description, the screen reader extracts (from the <src> element of the <img> tag) and reads the directory/filename.

Then it gives some examples. And the paper seems to raise the possibility that poor alt text and overlinking can cause extra clutter, so it would be wrong to focus the "obstacle" angle the way you want to, I think. It's more complicated than that. It also doesn't mention alt text (that I can see) in its basic suggestions, p. 20. SlimVirgin TALK contribs 16:55, 18 March 2010 (UTC)

What you call "some examples" are the two sub-bullets that followed the text that you quoted. I summarized them above; I just now bolded that part of the summary above, to call your attention to them again. The first sub-bullet are images that lack alt text, and which should have been marked as purely decorative with "|link=|alt=", but were not marked and so were read aloud as something like "link pix dot gee eye eff" when they should have been ignored by the screen reader. The second sub-bullet were images that lacked alt text but should have had it, and were read aloud as something like "link Wiktionary dash logo dash fifty-one pee ex" when they should have been "link Wiktionary". You are correct that there is a later bullet that talks about "confusing or meaningless names", which could indeed be talking about poorly written alt text but could also be talking about something else (the paper isn't 100% clear). However, this doesn't undermine the fact that the alt-text problems are listed first. And the paper itself says that this is a list of the "main problems" in access via a screen reader. If alt text is #1 on the list of the "main problems", and is given more space than any other problem, then what is inaccurate about calling lack of proper alt text an "important obstacle" toward blind access? Would you prefer it if we called it a "main problem" rather than an "important obstacle"? Anyway, I continue to fail to see why this important and directly-on-point citation was removed. Eubulides (talk) 06:32, 19 March 2010 (UTC)

Recent edits

I've tightened the text, removed some repetition and some of the examples. I've added the 7-20 word rule of thumb while making it clear that it depends on context; the previous version recommended that it be considerably shorter than 100. I've tried to tighten some of the recommended alt text examples, and made the flawed examples longer to illustrate the contrast. I've also moved a different image into the lead that shows at a glance how to write alt text. The page was 81 kilobytes and is down to 52. SlimVirgin TALK contribs 11:34, 18 March 2010 (UTC)

I like and support your changes. I find the whole thing much more palatable now and from what I read, all the suggestions and advice to make original research have been removed. All my policy concerns are addressed. —Charles Edward (Talk | Contribs) 12:35, 18 March 2010 (UTC)
Thanks, Charles. SlimVirgin TALK contribs 12:52, 18 March 2010 (UTC)
I support those changes. Good work. Regards, SunCreator (talk) 13:46, 18 March 2010 (UTC)

Some remaining issues

I'm confused about the advice given in Wikipedia:Alternative_text_for_images#Images_containing_text_and_unusual_characters. It recommends simply repeating unusual characters. For example, we should write "the character 袁", instead of something like "a character written in red on a white background". How is a screenreader going to know how to say 袁 and does this not contradict our advice not to use names unless they're well-known? And yet the final example recommends "A large overhead road sign with non-English words," and not "Picture of a large blue sign saying "Bienvenidos a ixmiquilpan: Corazon del Valle de Mezquital" and in smaller letters "hogä ehe nts'utk'ani"." Contradiction? SlimVirgin TALK contribs 12:57, 18 March 2010 (UTC)

That final example was changed recently (not by me), which leads to the apparent contradiction. The current example is obviously wrong. I would change it back to fix things, but I don't want to interrupt you as you're editing now. The intent is that alt text of an image that contains text should simply transcribe the text, and not interpret the text by translating it into English or anything like that (that's a job for the caption). A screen reader that sees "袁" will either say that it's an unknown character, or it will give the Unicode name of the character, or (if the screen reader is Chinese-aware) it will say the Chinese word. All of these behaviors are commonly encountered by people using screen readers and they will understand what's going on. Eubulides (talk) 16:21, 18 March 2010 (UTC)
I'll go into an old version and see what it said before the change. SlimVirgin TALK contribs 16:31, 18 March 2010 (UTC)
It looks as though it was you who added it back in August, in a section called tricky lettering. [10]. I think we should move it to talk for now and try to figure out what the best advice is. SlimVirgin TALK contribs 16:37, 18 March 2010 (UTC)
No, it was not I who changed the recommended alt text of the final example, which made it contradictory; it was SunCreator here, about three hours ago. Let's not delete the section simply because an editor made a mistake in it a few hours ago. Eubulides (talk) 16:58, 18 March 2010 (UTC)
I removed that we should describe video in detail and even place a full transcript on the file page, which I don't feel is reasonable. The section gave some long and detailed examples of alt text, and said: "Some readers are hearing-impaired and others use browsers without audio, so the alt text should not assume that the reader can hear a video. Prominent and brief words in the audio can be transcribed in the alt text, and it is helpful to provide a full transcript of the audio in the video's file description page." I tweaked the examples, so it now reads that the long version goes on the file page (if the editor wants to do that) and the brief description is the alt text. That section before and after. SlimVirgin TALK contribs 13:58, 18 March 2010 (UTC)
Yes, the intent was that the long version (if any) should go into the file page and the brief version in the alt text. I'll check the details later, after you're done editing. Eubulides (talk) 16:21, 18 March 2010 (UTC)

Tricky lettering

This section seems to contradict the "avoid tricky lettering" advice in the section below it, and also seems to be internally inconsistent, advising that we reproduce Mandarin and Hindi signs, but not non-English words. What do we want it to say exactly? SlimVirgin TALK contribs 16:42, 18 March 2010 (UTC)

Begins

When an image focuses on text, its alt text should normally contain that text even it contains unusual characters. Some screen readers cannot handle obscure Unicode characters, and either render them as "?" or ignore them entirely; therefore try to word alt text so that it makes some sense even if the unusual characters are ignored or read aloud as "?". In alt text for these images, omit details such as color or font unless the images focus on or illustrate these details.

Transcribe the image's text, even if it has unusual characters.
Context Image Flawed alt text Better alt text
First image of Yuan (surname)
The character "袁"
The character "Yuan" adapted from the 2nd century Yuan An stele
A character written in red on a white background The character "袁"
The Writing system section of Hindi
The word हिन्दी
"Hindi" in Devanagari
SVG image of a rendering for "Hindi" in Devanagari The word हिन्दी
The Orthography section of Otomi language
A large overhead road sign says "BIENVENIDOS A IXMIQUILPAN", then (smaller) "HOGÄ EHE NTS'U̲TK'ANI", then (larger again) "CORAZON DEL VALLE DEL MEZQUITAL".
Sign written in Otomi and Spanish in the Mezquital Valley
Picture of a large blue sign saying "Bienvenidos a ixmiquilpan: Corazon del Valle de Mezquital" and in smaller letters "hogä ehe nts'utk'ani". Behind the sign are rolling green hills. A large overhead road sign with non-English words.
What it should say
  • What we want it to say is what it used to say, until about 3 hours ago. The suggested text should transcribe the sign. These words are readable in the original context, but are not readable in WP:ALT, so we should also change the example to make this clearer. Something like this, perhaps:
Context Image Flawed alt text Better alt text
The Orthography section of Otomi language; in this context, the image is large enough so that the sign's text is readable
A large overhead road sign says "BIENVENIDOS A IXMIQUILPAN", then (smaller) "HOGÄ EHE NTS'U̲TK'ANI", then (larger again) "CORAZON DEL VALLE DEL MEZQUITAL".
Sign written in Otomi and Spanish in the Mezquital Valley
Picture of a large blue sign saying "Bienvenidos a ixmiquilpan: Corazon del Valle de Mezquital" and in smaller letters "hogä ehe nts'utk'ani". Behind the sign are rolling green hills. A large overhead road sign says "BIENVENIDOS A IXMIQUILPAN", then (smaller) "HOGÄ EHE NTS'U̲TK'ANI", then (larger again) "CORAZON DEL VALLE DEL MEZQUITAL".
  • This section should not contradict the "Avoid tricky lettering" advice. "Avoid tricky lettering" is supposed to say that you shouldn't use unusual symbols when you don't have to. This section is saying that sometimes you have to use unusual symbols, when the point of the image is to convey the unusual symbols.

Eubulides (talk) 16:58, 18 March 2010 (UTC)

Okay, thanks for fixing it. It's now internally consistent, but does it make sense to add unreadable symbols to alt text? SlimVirgin TALK contribs 17:30, 18 March 2010 (UTC)
Also, why do we spell out the words in the road sign, but not the symbols and not the URL? SlimVirgin TALK contribs 17:35, 18 March 2010 (UTC)

Alt text should not contain things hidden from sighted readers

The reason I removed the wording on the overhead road sign is that I cannot read it in WP:ALT without clicking to get the bigger version of the picture. Eubulides said before that "Alt text should not contain secrets hidden from sighted readers." I can however read it in Otomi_language#Orthography where it's of 233x310px size, but if we are showing it as a good example of alt text I suggest we show the picture in the same size not a mini(90x120px) version of it. Regards, SunCreator (talk) 19:20, 18 March 2010 (UTC)

I don't see a problem with having to click on it to see extra detail, Sun. SlimVirgin TALK contribs 19:26, 18 March 2010 (UTC)
You could be right. Pondering that at the moment. Regards, SunCreator (talk) 19:39, 18 March 2010 (UTC)

Fangs

I downloaded this to give me an idea of what a screen reader would say about the intro to Death of Ian Tomlinson. This is what it gave me for the lead, and bear in mind that, being the lead, this has only four references (a ref produces "This page link left bracket four right bracket").

Extended content

Page has one frame, thirty-eight headings and seven hundred forty linksDeath of Ian Tomlinson dash Wikipedia, the free encyclopedia dash Internet ExplorerHeading level one left bracket Link edit right bracket Death of Ian TomlinsonHeading level three A Link featured article from Wikipedia, the free encyclopedia left paren Redirected from Link Ian Tomlinson right paren quote Ian Tomlinson quote redirects here. For the Australian athlete, see Link Ian Tomlinson left paren athlete right paren .Table caption colon Death of Ian TomlinsonTable with one column and ten rowsLinkGraphicA small crowd scene. On the right, four people dressed in uniform, their heads and face mostly not visible, wearing yellow and blue jackets, black trousers and black shoes. They are carrying long thin sticks, and round transparent shields. On the left, there are three men. One is on the ground, sitting with his legs straight out, and his arms raised, looking at the people dressed in uniform. He is wearing a grey and blue top and black trousers with a white stripe. Two men are leaning over him; one is holding his arms. The latter is hearing a dark hooded top, grey trousers with white stripes, and has a grey and blue bag over his shoulders. Ian Tomlinson remonstrates with police after being pushed to the ground, minutes before he died. Date one April two thousand nine Time c. nineteen colon thirty hours Link GMT LocationLink Cornhill , Link City of London First reporter Paul Lewis, Link The Guardian Filmed by American investment fund manager Charges None Awards Bevins Prize for outstanding investigative journalism for Paul LewisThis page link left bracket one right bracket FootageLink American businessman's video obtained by The GuardianTable end Ian Tomlinson left paren seven February one thousand nine hundred sixty-two – one April two thousand nine right paren was an English newspaper vendor who collapsed and died in the Link City of London on his way home from work during the Link G dash twenty summit protests . A first Link postmortem examination indicated that he had suffered a heart attack brought on by coronary artery disease, and had died of natural causes.This page link left bracket two right bracket His death became controversial a week later when Link The Guardian obtained footage of his last moments, filmed by an American investment fund manager who was visiting London. The video shows Tomlinson being struck on the leg from behind by a police officer wielding an Link expandable baton , then pushed to the ground by the same officer. It appears to show no provocation on Tomlinson's part—he was not a protester, and at the time he was struck, the footage shows him walking along with his hands in his pockets. He walked away after the incident, but collapsed and died moments later.This page link left bracket three right bracket After The Guardian published the video, the Link Independent Police Complaints Commission left paren IPCC right paren began a criminal inquiry from which the police were removed. A second postmortem, ordered by the IPCC and Tomlinson's family, indicated that Tomlinson had died from an abdominal Link haemorrhage , the cause of which remains unknown. The officer in question, who has not been named, was interviewed on suspicion of Link manslaughter , and a third postmortem was conducted at the request of his defence team, the results of which have not been released. The IPCC completed its investigation in August two thousand nine and passed its file to the Link Crown Prosecution Service . No charges have been announced.This page link left bracket four right bracket Tomlinson's death sparked an intense debate in the UK about what appeared to be a deteriorating relationship between the police and the public, the degree to which the IPCC is independent of the police, and the role of citizens in monitoring police and government activity—so dash called Link sousveillance . There was criticism of the news coverage too, the mayor of London, Link Boris Johnson , calling it quote an orgy of cop bashing. quote The incident was compared to previous deaths involving either police contact or allegedly inadequate investigations, such as the deaths of Link Blair Peach left paren one thousand nine hundred seventy-nine right paren , Link Stephen Lawrence left paren one thousand nine hundred ninety-three right paren , and Link Jean Charles de Menezes left paren two thousand five right paren , each of which acted as a watershed in the public's perception of policing in the country.This page link left bracket five right bracket In response to public concerns, the Link Chief Inspector of Constabulary , Link Denis O'Connor , published a one hundred fifty dash page report in November two thousand nine that aimed to restore Britain's traditional consent dash based model of policing. The Guardian hailed the report as a blueprint for wholesale reform.

If this is what's being read out, the lack of alt text is just one of a number of problems, links and references being the most obvious. SlimVirgin TALK contribs 19:43, 18 March 2010 (UTC)

I went to a version of a page I'd collected before for being over-referenced: Gaza War, the lead from last July.
Extended content

The Gaza War, part of the ongoing left bracket left bracket Israeli–Palestinian conflict right bracket right bracket , less ref greater left bracket http colon slash slash www.csis.org slash media slash csis slash pubs slash zero underline gaza underline war.pdf Anthony H. Cordesman, ‘THE “GAZA WAR” colon A Strategic Analysis,’ Center for Strategic and International Studies, February two thousand nine p point five right bracket less slash ref greater commenced when left bracket left bracket Israel right bracket right bracket launched a left bracket left bracket military campaign right bracket right bracket in the left bracket left bracket Gaza Strip right bracket right bracket on December twenty-seven two thousand eight codenamed Operation Cast Lead left paren left brace left brace lang dash he vertical bar מבצע עופרת יצוקה right brace right brace right paren , less ref greater left brace left brace cite news vertical bar url equals http colon slash slash www.haaretz.com slash hasen slash spages slash one million fifty thousand four hundred five .html vertical bar title equals ANALYSIS slash IAF strike on Gaza is Israel’s version of ‘shock and awe’ vertical bar last equals Harel vertical bar first equals Amos vertical bar date equals December twenty-seven two thousand eight vertical bar publisher equals Ha’aretz vertical bar accessdate equals December twenty-seven two thousand eight right brace right brace less slash ref greater with the stated aim of stopping left bracket left bracket Hamas right bracket right bracket rocket attacks on southern Israel and arms smuggling into Gaza. less ref name equals quote APboat quote greater left brace left brace cite web vertical bar url equals http colon slash slash www.google.com slash hostednews slash ap slash article slash ALeqMfive ioi underline zerojtOnine RjMwPNRoXNCndRPRqthree gDninety-seven HIUzeroOzero vertical bar title equals Gaza boat explodes in apparent attack, no injuries vertical bar last equals Weizman vertical bar first equals Steve vertical bar date equals April thirteen two thousand nine vertical bar publisher equals Assciated Press vertical bar accessdate equals two thousand nine dash four dash twenty-five right brace right brace less slash ref greater less ref greater left bracket http colon slash slash www.mfa.gov.il slash MFA slash Terrorism dash plus Obstacle plus to plus Peace slash Hamas plus war plus against plus Israel slash Israel underline strikes underline back underline against underline Hamas underline terror underline infrastructure underline Gaza underline twenty-seven dash Dec dash two thousand eight .htm quote Israel strikes back against Hamas terror infrastructure in Gaza twenty-seven dash Dec dash two thousand eight . quote right bracket left bracket left bracket Israel Ministry of Foreign Affairs right bracket right bracket . January twenty-one two thousand nine . February two two thousand nine . less slash ref greater The conflict has also been called the Gaza massacre in the left bracket left bracket Arab world right bracket right bracket left paren left brace left brace lang dash ar vertical bar مجزرة غزة right brace right brace right paren . less ref greater left brace left brace Cite news vertical bar url equals http colon slash slash www.turkishweekly.net slash news slash sixty-two thousand five hundred forty-three slash factions dash refuse dash abbas dash thirty-nine dash call dash for dash unity dash meeting dash amid dash gaza dash massacre.html vertical bar archiveurl equals http colon slash slash www.webcitation.org slash five djtfPninety-two i vertical bar archivedate equals two thousand nine dash one dash eleven right brace right brace star left brace left brace Cite news vertical bar title equals Iraqi leaders discuss Gaza massacre vertical bar url equals http colon slash slash www.gulfnews.com slash region slash Iraq slash ten million two hundred seventy thousand seven hundred sixty-one .html vertical bar date equals two thousand eight dash twelve dash twenty-eight vertical bar accessdate equals two thousand nine dash one dash zero vertical bar publisher equals gulfnews.com vertical bar archiveurl equals http colon slash slash www.webcitation.org slash five dfWone Ceight nU vertical bar archivedate equals two thousand nine dash one dash zero right brace right brace star left brace left brace Cite news vertical bar title equals Israel airstrikes on Gaza kill at least two hundred twenty-five vertical bar url equals http colon slash slash www.khaleejtimes.com slash DisplayArticlezero.asp?xfile equals data slash middleeast slash two thousand eight slash December slash middleeast underline Decemberfive hundred seven .xml and section equals middleeast vertical bar agency equals left bracket left bracket Deutsche Presse dash Agentur vertical bar Deutsche Presse dash Agentur left paren DPA right paren right bracket right bracket vertical bar publisher equals Khaleej Times vertical bar date equals two thousand eight dash twelve dash twenty-seven vertical bar accessdate equals two thousand nine dash one dash eleven vertical bar archiveurl equals http colon slash slash www.webcitation.org slash five eULZTBGU vertical bar archivedate equals two thousand nine dash two dash ten right brace right brace star left brace left brace Cite news vertical bar title equals Hamas denies firing rockets from Lebanon vertical bar url equals http colon slash slash www.sbs.com.au slash news slash article slash one million four thousand five hundred twenty-six slash Hamas dash denies dash firing dash rockets dash from dash Lebanon vertical bar agency equals left bracket left bracket Agence France dash Presse right bracket right bracket vertical bar publisher equals left bracket left bracket Special Broadcasting Service right bracket right bracket vertical bar date equals two thousand nine dash one dash zero vertical bar accessdate equals two thousand nine dash one dash eleven vertical bar archiveurl equals http colon slash slash www.webcitation.org slash five djdLfRthirty-one vertical bar archivedate equals two thousand nine dash one dash eleven right brace right brace star quote Diplomatic race to stop the Gaza massacre quote dash left brace left brace Cite news vertical bar title equals سباق دبلوماسي لوقف مذبحة غزة vertical bar url equals http colon slash slash news.bbc.co.uk slash hi slash arabic slash middle underline east underline news slash newsid underline seven million eight hundred ten thousand slash seven million eight hundred ten thousand nine hundred sixty-eight .stm vertical bar agency equals left bracket left bracket BBC right bracket right bracket Arabic vertical bar date equals two thousand nine dash one dash five vertical bar accessdate equals two thousand nine dash one dash eleven vertical bar archiveurl equals http colon slash slash www.webcitation.org slash five djZbRLwy vertical bar archivedate equals two thousand nine dash one dash eleven right brace right brace star quote Victims of the Gaza massacre of nearly three thousand in the ninth day of the aggression quote dash left brace left brace Cite news vertical bar title equals ضحايا مجزرة غزة يقارب الـthree آلاف في اليوم التاسع للعدوان vertical bar url equals http colon slash slash www.moheet.com slash show underline files.aspx?fid equals two hundred seven thousand seven hundred sixty vertical bar date equals two thousand nine dash one dash four vertical bar accessdate equals two thousand nine dash two dash zero vertical bar agency equals Arab News Network left paren Moheet right paren vertical bar archiveurl equals http colon slash slash www.webcitation.org slash five ddnmnine cIU vertical bar archivedate equals two thousand nine dash one dash six right brace right brace star quote Gaza massacre roused anti dash Semitism in Europe quote dash left brace left brace Cite news vertical bar title equals مجزرة غزة أيقظت اللاسامية في أوروبا vertical bar url equals http colon slash slash www.daralhayat.com slash opinion slash two dash two thousand nine slash Article dash twenty million ninety thousand two hundred six dash four csix bfive ethree b dash czeroaeight dash ten ed dash fourteen d dash three hundred four thousand six hundred ninety-three dsixty-nine dfifty-five slash story.html vertical bar date equals two thousand nine dash two dash seven vertical bar accessdate equals two thousand nine dash two dash zero vertical bar agency equals Dar al hayat vertical bar archiveurl equals http colon slash slash www.webcitation.org slash five eQJSsix Pfive i vertical bar archivedate equals two thousand nine dash two dash seven right brace right brace less slash ref greater less nowiki greater less ! dash dash Basic outline of the conflict, consensus version, please keep compact! dash dash greater A left bracket left bracket two thousand eight Israel–Hamas ceasefire vertical bar fragile six dash month truce between Hamas and Israel right bracket right bracket expired on December nineteen two thousand eight . less ref greater left brace left brace cite news vertical bar title equals TIMELINE dash Israeli dash Hamas violence since truce ended vertical bar url equals http colon slash slash uk.reuters.com slash article slash topNews slash idUKTREfifty Tera four hundred twenty-three billion three hundred twenty million ninety thousand one hundred five vertical bar publisher equals Reuters right brace right brace less slash ref greater Hamas and Israel could not agree on conditions to extend the truce. less ref greater left brace left brace cite news vertical bar url equals http colon slash slash news.bbc.co.uk slash two slash hi slash middle underline east slash seven million seven hundred ninety-seven thousand one hundred forty-four .stm vertical bar publisher equals BBC vertical bar title equals Hamas 'might renew' truce in Gaza right brace right brace less slash ref greater less ref greater left brace left brace cite news vertical bar url equals http colon slash slash www.huffingtonpost.com slash two thousand nine slash one slash zero slash israel dash rejected dash hamas dash cea underline n underline one hundred fifty-six thousand six hundred thirty-nine .html?page equals two and show underline comment underline id equals nineteen million five hundred fifty-eight thousand eight hundred eighty-eight number comment underline nineteen million five hundred fifty-eight thousand eight hundred eighty-eight vertical bar publisher equals Huffington Post vertical bar title equals Israel Rejected Hamas Ceasefire Offer In December right brace right brace less slash ref greater Hamas blamed Israel for not lifting the left bracket left bracket Gaza Strip blockade right bracket right bracket , and for an Israeli raid on a purported tunnel, crossing the border into the Gaza Strip from Israel on November four less ref greater left bracket http colon slash slash www.csis.org slash media slash csis slash pubs slash zero underline gaza underline war.pdf Anthony H. Cordesman, ‘THE “GAZA WAR” colon A Strategic Analysis,’ Center for Strategic and International Studies, February two thousand nine right bracket p point nine less slash ref greater which it held constituted a serious breach of the truce. less ref greater left bracket http colon slash slash www.foxnews.com slash story slash two .html ‘Israeli Airstrike on Gaza Threatens Truce with Hamas,’ Fox News, November four two thousand eight right bracket less slash ref greater less ref name equals Guardiantwenty million ninety-one thousand one hundred five slash greater Israel accused Hamas of violating the truce, citing the frequent left bracket left bracket List of rocket and mortar attacks in Israel in two thousand eight vertical bar rocket and mortar attacks right bracket right bracket on Israeli cities. less ref greater left brace left brace cite news vertical bar author equals Ibrahim Barzak vertical bar coauthors equals Amy Teibel vertical bar title equals World leaders converge on Israel in push for truce vertical bar url equals http colon slash slash pressherald.mainetoday.com slash story.php?id equals two hundred thirty-one thousand three hundred thirty-four and ac equals PHnws vertical bar date equals two thousand nine dash one dash zeros vertical bar publisher equals Maine Sunday Telegram right brace right brace less slash ref greater The Israeli operation began with an left bracket left bracket Timeline of the two thousand eight –two thousand nine Israel–Gaza conflict number twenty-seven December two thousand eight vertical bar intense bombardment right bracket right bracket of the left bracket left bracket Gaza Strip right bracket right bracket , less ref name equals quote Times Online two thousand eight dash twelve dash twenty-eight dash b quote greater http colon slash slash www.timesonline.co.uk slash tol slash news slash world slash middle underline east slash articlefive million four hundred seven thousand three hundred eighty-two .ece less slash ref greater targeting Hamas bases, police training camps, less ref greater left bracket http colon slash slash www.iht.com slash articles slash two thousand eight slash twelve slash twenty-seven slash africa slash twenty-eight mideast underline update.php Israeli Gaza strike kills more than two hundred right bracket , left bracket left bracket International Herald Tribune right bracket right bracket , two thousand eight dash twelve dash twenty-seven . less slash ref greater police headquarters and offices. less ref greater left brace left brace cite news vertical bar url equals http colon slash slash www.haaretz.com slash hasen slash spages slash one million fifty thousand four hundred thirty-two .html vertical bar title equals Most Hamas bases destroyed in four minutes vertical bar author equals Amos Harel vertical bar publisher equals left bracket left bracket Haaretz right bracket right bracket vertical bar accessdate equals December twenty-eight two thousand eight right brace right brace less slash ref greater less ref greater left brace left brace cite news vertical bar url equals http colon slash slash www.jpost.com slash servlet slash Satellite?cid equals one Tera two hundred thirty billion one hundred eleven million seven hundred fourteen thousand nine hundred sixty-nine and pagename equals JPost percent two FJPArticle percent two FShowFull vertical bar title equals A year's intel gathering yields 'alpha hits' vertical bar author equals Yaakov Katz vertical bar publisher equals left bracket left bracket Jerusalem Post right bracket right bracket vertical bar accessdate equals December twenty-eight two thousand eight right brace right brace less slash ref greater left bracket left bracket Incidents in the two thousand eight –two thousand nine Israel–Gaza conflict number civilian infrastructure vertical bar Civilian infrastructure right bracket right bracket , including left bracket left bracket mosques right bracket right bracket , houses, medical facilities and schools, were also attacked, as Israel stated that many of them were used by combatants, and as storage spaces for weapons and rockets. less ref name equals quote guardian dash israel dash gaza dash attacks quote greater left brace left brace cite news vertical bar url equals http colon slash slash www.guardian.co.uk slash world slash two thousand nine slash jan slash two slash israel dash gaza dash attacks vertical bar title equals Israeli warplanes destroy Gaza houses and mosque as air strikes continue vertical bar last equals McCarthy vertical bar first equals Rory vertical bar coauthors equals David Batty and agencies vertical bar date equals January two two thousand nine vertical bar publisher equals Guardian vertical bar language equals English vertical bar accessdate equals two thousand nine dash one dash five right brace right brace less slash ref greater Hamas intensified its rocket and mortar attacks against targets in Israel throughout the conflict, hitting previously untargeted cities such as left bracket left bracket Beersheba right bracket right bracket and left bracket left bracket Ashdod right bracket right bracket . less ref greater left brace left brace cite news vertical bar url equals http colon slash slash www.ynetnews.com slash articles slash three thousand eight hundred eight L dash three hundred sixty-four million five hundred forty-eight thousand five hundred .html vertical bar title equals Rockets land east of Ashdod vertical bar date equals December twenty-eight two thousand eight vertical bar publisher equals Ynetnews vertical bar accessdate equals December twenty-eight two thousand eight right brace right brace less slash ref greater less ref name equals ynet dash school dash rocket greater left brace left brace cite web vertical bar title equals Rockets reach Beersheba, cause damage vertical bar url equals http colon slash slash www.ynetnews.com slash articles slash three thousand eight hundred eight L dash three hundred sixty-four million seven hundred fifty-six thousand nine hundred .html vertical bar publisher equals left bracket left bracket YNET right bracket right bracket vertical bar date equals two thousand eight dash twelve dash thirty vertical bar language equals English right brace right brace less slash ref greater On January three two thousand nine the Israeli ground invasion began. less ref name equals msnbc dash ground dash invasion greater left brace left brace cite news vertical bar url equals http colon slash slash www.msnbc.msn.com slash id slash twenty-eight million four hundred eighty-three thousand seven hundred fifty-six slash vertical bar title equals Israel Confirms Ground Invasion Has Started vertical bar date equals two thousand nine dash one dash three vertical bar publisher equals left bracket left bracket MSNBC right bracket right bracket vertical bar accessdate equals two thousand nine dash one dash four right brace right brace left brace left brace Dead link vertical bar date equals February two thousand nine right brace right brace less slash ref greater less ref name equals quote ap dash ground dash attack quote greater left brace left brace cite news vertical bar url equals http colon slash slash abcnews.go.com slash International slash wireStory?id equals six million five hundred sixty-eight thousand nine hundred sixty-one vertical bar title equals Israeli ground troops invade Gaza to halt rockets vertical bar last equals BARZAK vertical bar first equals IBRAHIM vertical bar coauthors equals JASON KEYSER vertical bar date equals two thousand nine dash one dash four vertical bar agency equals left bracket left bracket Associated Press right bracket right bracket vertical bar language equals English vertical bar accessdate equals two thousand nine dash one dash four right brace right brace less slash ref greater less ! dash dash Summarise quote Reactions quote section, and humanitarian situation. dash dash greater left bracket left bracket International reaction to the two thousand eight –two thousand nine Israel–Gaza conflict vertical bar International reactions during the conflict right bracket right bracket included calls for an immediate ceasefire as in the left bracket left bracket United Nations Security Council Resolution one thousand eight hundred sixty right bracket right bracket , and concern about the left bracket left bracket humanitarian aid vertical bar humanitarian right bracket right bracket situation in the Gaza Strip and the hindrances in delivering aid. less ref name equals quote reuterstruce quote greater left bracket http colon slash slash www.reuters.com slash article slash newsOne slash idUSLSsixty-nine Tera three hundred ninety-one billion six hundred twenty million eighty-one thousand two hundred thirty Israel and Hamas under pressure for Gaza aid truce right bracket Reuters two thousand eight dash twelve dash thirty less slash ref greater less ! dash dash War Crimes dash dash greater Human rights groups and aid organisations have accused Hamas and Israel of left bracket left bracket war crime right bracket right bracket s and called for independent investigations and lawsuits. less ref name equals hs greater left brace left brace cite news vertical bar title equals Gaza 'human shields' criticised vertical bar url equals http colon slash slash news.bbc.co.uk slash one slash hi slash world slash middle underline east slash seven million eight hundred eighteen thousand one hundred twenty-two .stm right brace right brace less slash ref greater less ref name equals quote btselem quote greater left bracket http colon slash slash www.btselem.org slash english slash Press underline Releases slash twenty million seventy thousand six hundred twenty-five .asp quote Holding Gilad Shalit as a hostage is a war crime quote right bracket dash B'Tselem press release, issued twenty-five June two thousand seven less slash ref greater less ref greater left brace left brace cite web vertical bar url equals http colon slash slash www.guardian.co.uk slash world slash two thousand nine slash jan slash thirteen slash gaza dash israel dash war dash crimes vertical bar title equals Demands grow for Gaza war crimes investigation vertical bar last equals McGreal vertical bar first equals Chris vertical bar date equals two thousand nine dash one dash twenty-three vertical bar publisher equals The Guardian vertical bar accessdate equals two thousand nine dash one dash twenty-three right brace right brace less slash ref greater less ! dash dash Casualties and Destruction dash dash greater Between one thousand one hundred sixty-six to one thousand four hundred seventeen Palestinians and thirteen Israelis died in the conflict. less ref name equals quote reuterstwo thousand nine jantwelve quote greater left bracket http colon slash slash www.newsdaily.com slash stories slash trefive thousand fifty-three rseven dash us dash palestinians dash israel slash Israel tightens grip on urban parts of Gaza right bracket . By Nidal al dash Mughrabi. January twelve two thousand nine . left bracket left bracket Reuters right bracket right bracket . less slash ref greater less ref name equals quote idf underline updated underline death underline toll quote greater left brace left brace cite news vertical bar url equals http colon slash slash www.jpost.com slash servlet slash Satellite?cid equals one Tera two hundred thirty-seven billion seven hundred twenty-seven million five hundred fifty-two thousand fifty-four and pagename equals JPost percent two FJPArticle percent two FShowFull vertical bar title equals IDF releases Cast Lead casualty numbers vertical bar date equals two thousand nine dash three dash twenty-six vertical bar publisher equals JPost vertical bar accessdate equals two thousand nine dash three dash twenty-six vertical bar last equals Lappin vertical bar first equals Yaakov right brace right brace less slash ref greater less ref name equals quote PCHR underline CIV underline STAT quote greater left brace left brace cite web vertical bar url equals http colon slash slash www.cbsnews.com slash stories slash two thousand nine slash one slash twenty-two slash world slash mainfour million seven hundred forty-six thousand two hundred twenty-four .shtml?source equals RSSattr equals World underline four million seven hundred forty-six thousand two hundred twenty-four vertical bar title equals Rights Group Puts Gaza Death Toll At one thousand two hundred eighty-four vertical bar last equals Younis vertical bar first equals Khan vertical bar publisher equals CBS vertical bar accessdate equals two thousand nine dash two dash seventeen right brace right brace less slash ref greater The number of combatant and non dash combatant casualties is a subject of ongoing contention. less ref name equals bbc dash count dash contro greater left brace left brace cite news vertical bar url equals http colon slash slash news.bbc.co.uk slash two slash hi slash middle underline east slash seven million eight hundred fifty-five thousand seventy .stm vertical bar title equals Counting casualties of Gaza's war vertical bar last equals Bell vertical bar first equals Bethany vertical bar date equals January twenty-eight two thousand nine vertical bar publisher equals BBC vertical bar accessdate equals two thousand nine dash two dash seventeen right brace right brace less slash ref greater Casualty figures have been difficult to verify due to the limited amount of journalists allowed in Gaza during the conflict. less ref name equals bbc dash count dash contro slash greater In the days following the ceasefire, the BBC reported that more than four hundred thousand Gazans were left without running water. less ref name equals quote bbcjanone hundred ninety-two thousand nine dash earthquate quote greater left bracket http colon slash slash news.bbc.co.uk slash two slash hi slash middle underline east slash seven million eight hundred thirty-eight thousand six hundred eighteen .stm Gaza 'looks like earthquake zone' right bracket . left bracket left bracket BBC News right bracket right bracket . January nineteen two thousand nine . less slash ref greater The BBC further reported that four thousand homes had been ruined, leaving tens of thousands of people homeless. less ref name equals quote BBC vertical bar Scale of Gaza destruction emerges quote greater left bracket http colon slash slash news.bbc.co.uk slash one slash hi slash world slash middle underline east slash seven million eight hundred thirty-six thousand eight hundred sixty-nine .stm 'Scale of Gaza destruction emerges' right bracket . left bracket left bracket BBC News right bracket right bracket . January nineteen two thousand nine . less slash ref greater less ! dash dash Ceasefire dash dash greater The conflict came to an end on January eighteen after first Israel and then Hamas announced left bracket left bracket unilateral right bracket right bracket ceasefires. less ref name equals quote CNN Gaza quote greater left brace left brace cite web vertical bar url equals http colon slash slash edition.cnn.com slash two thousand nine slash WORLD slash meast slash one slash eighteen slash israel.gaza slash index.html vertical bar title equals Hamas, Israel set independent cease dash fires vertical bar publisher equals CNN International right brace right brace less slash ref greater less ref name equals quote Journalist Enters Gaza quote greater left brace left brace cite news vertical bar title equals Israel wants rapid Gaza pullout vertical bar publisher equals BBC vertical bar date equals January eighteen two thousand nine vertical bar url equals http colon slash slash news.bbc.co.uk slash two slash hi slash middle underline east slash seven million eight hundred thirty-six thousand six hundred sixty .stm right brace right brace less slash ref greater On January twenty-one Israel completed its withdrawal from the Gaza Strip. less ref greater http colon slash slash news.bbc.co.uk slash one slash hi slash world slash middle underline east slash seven million eight hundred forty-one thousand nine hundred two .stm less slash ref greater On March two it was reported that international donors had pledged left bracket left bracket USD vertical bar dollars four point five billion right bracket right bracket in aid for the Palestinians, mainly for rebuilding Gaza after Israel's offensive. less ref greater left brace left brace cite web vertical bar url equals http colon slash slash news.bbc.co.uk slash one slash hi slash world slash middle underline east slash seven million nine hundred eighteen thousand one hundred five .stm vertical bar title equals Billions pledged to rebuild Gaza vertical bar date equals two thousand nine dash three dash two vertical bar publisher equals left bracket left bracket BBC News right bracket right bracket vertical bar accessdate equals two thousand nine dash three dash four right brace right brace less slash ref greater

It would drive anyone mad to have to listen to that. :) SlimVirgin TALK contribs 19:51, 18 March 2010 (UTC)
The only solution to this is if we have separate user-friendly alternate version of this page for screen readers without the links, footnotes and such other complicating issues. WikiProject Spoken Wikipedia faces such issues while converting an article into a Spoken Wikipedia script. Maybe a bot could strip out the screen-reader unfriendly elements. AshLin (talk) 06:23, 19 March 2010 (UTC)
  • SlimVirgin is of course correct that Wikipedia has too many links, and that it formats reference links poorly. I would like to fix this, and have proposed one obvious fix for the reference links (namely, omit the square brackets; they're really not needed), but this proposal has not been accepted. Further improvement in this area would be greatly helpful, and if takes a separate bot or whatever I hope someone with more expertise than I can write it. However, the bot (or whatever) can't be expected to supply alt text when alt text is missing, so we need to do that part by hand.
  • Although Fangs is useful to give sighted readers a first approximation to how blind readers would hear a page, the transcriptions above are misleading in that they completely fail to capture crucial parts of the overall experience. Blind readers hardly ever sit quietly and listen to a page, or even that much of a page. Instead, they jump around the page, using TAB and other keys, and typically skip a lot of the text. It is quite easy and common for them to tab to the next link, or to the next element in the list, or so forth, and yet these transcriptions don't show the places that one can tab to.
  • This is one of the reasons, by the way, that it's not vital for alt text to be extremely terse. What's vital is for the first few words of the alt text to be the important words. This is because users of screen readers will listen to the first few words of alt text, decide whether they want to hear the whole thing, and when they don't (which is fairly common) they'll press a key and will immediately skip to the next part of the article, or to the next section, or to the next link, or whatever. It doesn't really bother them much if the alt text is too long. What bothers them is if it's missing, or misleading, or omits important information.
  • Although the bad formatting and overlinking generate some useless words, the missing alt text causes the reader to lose important information about the article. Omitting important information is a far more serious obstacle than adding a few noise words. And this is why, in practice, alt text issues are among the biggest barriers to blind access to Wikipedia.

Eubulides (talk) 07:00, 19 March 2010 (UTC)

Could some of these issues be resolved with a different Preferences Skin? Failing that, I would have thought it would be possible to render the Wikitext differently for accessibility, though that amount of work would probably require funding and a grant from somewhere. Colin°Talk 08:28, 19 March 2010 (UTC)
What I'm thinking is that we need to set up a specialist team or wikiproject to look into this. We don't know what kind of alt text is best, and the way we write it seems to have no consistent governing principles. What would be great is if we could get feedback from a group of visually impaired users. We need to know what kind of alt text paints a picture for them. Just as an example, Sun Creator removed that the London townhouse had three or four sash windows per floor. On the one hand he's right; it sounds like clutter. On the other hand, it gives the reader a sense of its width and how imposing it is, because many such houses in London have only two per floor. It would be good to hear from visually impaired people how much that kind of detail matters to them. We also need to build up a set of good published sources so we can follow their advice.
If that can be achieved, it would be great if a team of editors would emerge who were willing to become specialist alt-text writers. Whereas it's fine for Google to say all images must have alt text—all they have to do is hire people to write it—it's not easy for us to impose on editors who are already stretched. Also, because we use non-specialist volunteers, the quality of alt text we're adding may be cluttering the screen readers up, rather than helping. SlimVirgin TALK contribs 08:43, 19 March 2010 (UTC)
Fangs actually displays some colour-coded information that was lost in copying. Therefore I guess that with a bit more experience, in particular having seen a real screenreader in action, operated by a competent user, it can be an excellent tool.
Regarding the brackets in references, for my account they are removed. I don't remember what I did to achieve this: I can't find the option in my preferences or the script in my monobook.js. Hans Adler 09:06, 19 March 2010 (UTC)
The only colour-coding I can see is for the word "Link." Here's a sentence from Philip Larkin in Fangs:

His first book of poetry, The North Ship , was published in one thousand nine hundred forty-five followed by two novels, Jill left paren one thousand nine hundred forty-six right paren and A Girl in Winter left paren one thousand nine hundred forty-seven right paren , but he came to prominence in one thousand nine hundred fifty-five with the publication of his second collection of poems, Link The Less Deceived , followed by Link The Whitsun Weddings left paren one thousand nine hundred sixty-four right paren and Link High Windows left paren one thousand nine hundred seventy-four right paren .

Even with each example of "Link" in grey, it would still be pretty unpleasant to listen to. Hans, what do I need to do to get a real screen-reading experience? I'd like to try it. SlimVirgin TALK contribs 10:22, 19 March 2010 (UTC)
Unfortunately I can't help. I don't know anybody who is blind. I guess one would have to install one that is widely used and practise with it for several hours. But that's just common sense.
You are right about the colours. The section headings are marked blue, but that's irrelevant for you example. Hans Adler 11:21, 19 March 2010 (UTC)
I'm blind, I use the screen reader JAWS, and I've made thousands of edits on Wikipedia. I use fairly eccentric settings for JAWS (especially regarding punctuation and how much text is spoken). I've gotten used to the links, but while trying to show Wikipedia to a friend of mine who was a new computer user, I noticed that she had a lot of trouble with them. I know several computer-literate blind people who often use Wikipedia, so it can't be *that* hard. Regarding the brackets, I never notice them because of my punctuation settings. I set my punctuation to "some", but on any higher settings, the brackets are always read. The default is either "most" or "all"; I can't remember which one it is, but it's immaterial because the brackets in footnotes would always be read for users of JAWS who use default settings. The only thing that Fangs gets wrong is the way years are spoken; they are spoken like "nineteen sixty two", as in normal speech, rather than "one thousand nine hundred sixty two". Anyone can download and use a demo version of JAWS at this link. Graham87 05:31, 20 March 2010 (UTC)

References available off line.

Some relevant quotes from scholarly articles. Full-text copies available if you send me an email.

  • "Enabling or disabling technologies? A critical approach to web accessibility Alison Adam and David Kreps Information Systems, Organisations and Society Research Centre [www.emeraldinsight.com/0959-3845.htm]. "The “semantic” information must be available to the assistive software separately from the graphical representation made available to sighted users of the web. This, unfortunately, is all too often forgotten by an industry wedded to the “graphical paradigm”. The IMG element of HTML, for example, is used to place an image on a web page. The ALT attribute of this element was introduced in HTML 2, for web authors to provide a text equivalent for images. The RNIB (UK based Royal National Institute for the Blind) recommend five words, e.g. ALT ¼ “dog leaps for a stick”. Speech synthesis software reads the ALT text. (Automated checking software will accept ALT ¼ “image.jpg” in the code, as a valid ALT attribute. Not exactly helpful.) If the dog is leaping for a stick over a canal, and the paragraph of text below the picture is about how funny it is that the dog gets wet, it may be necessary to describe the picture in more detail. The LONGDESC attribute of the IMG element, introduced in HTML4, allows web authors to provide the URL of a page where a longer description may be found."
  • "An empirical investigation into the difficulties experienced by visually impaired Internet users" Emma Murphy Æ Ravi Kuber Æ Graham McAllister Æ Philip Strain Æ Wai Yu. Univ Access Inf Soc (2008) 7:79–91."People who had vision earlier-on in their lives or who were partially sighted seemed to be more interested in images. They mentioned that some diagrams could play an important role on the Web, often supplementing the text presented on the page. Half of the participants felt that they were missing out on a perceptual experience which they thought that the fully sighted community experience when viewing images."..."Sites were generally found to lack alternatives to graphics and images, such as alternative text (ALT text) descriptions.... Blind users wanted short and succinct alternative text for larger images. According to blind participants, small spacer images should not be labelled with ALT text, as it was frustrating trying to work out if the image was for decorative purposes or if the images had intended meaning for the page. For congenitally blind users, who have never seen an image, a textual description may be difficult to visualise in their mind. Therefore, some blind users were uninterested in the mention of colours or visual descriptions, which meant very little to them. One user recommended making ALT text into a hyperlink, allowing users to have the option of listening to the description, if they wanted more information about the image. By manipulating the verbosity of the text in this link, users could gain as much information as they wanted."--Slp1 (talk) 18:34, 19 March 2010 (UTC)
Thanks for this. One bit puzzles me; what does this mean: By manipulating the verbosity of the text in this link, users could gain as much information as they wanted? Carcharoth (talk) 21:09, 19 March 2010 (UTC)
I took that to be an oblique reference to the "alt" vs "longdesc" distinction. --Malleus Fatuorum 21:15, 19 March 2010 (UTC)

Heart of the matter

Colin made a very good point above, that is perhaps really getting to the heart of all the disagreement here. He said "... we are singing from the same hymnsheet now but still tripping up on terminology. Could I suggest we don't use the abbreviation 'alt' unless we mean the wiki markup. And probably avoid referring to the HTML markup unless we are having a technical discusion."

For our purposes, the HTML markup is completely irrelevant, as none of us writes it; what we do is to fill in the boxes for the {{Image}} template, which generates the HTML alt attribute in a way that could very easily be changed should we decide that's necessary.

  • Web accessibility guidelines mandate an HTML alt text for every image, even if that's a default alt="". All wikipedia images have a default alt text, because that's generated by the {{Image}} template}}.
  • If no alt attribute is specified in the {{Image}} template, then the template appears to default to the image's file name; that could easily be changed to something else, like "see caption".
  • That all the images we're concerned with will have captions further muddies the water, because the caption may often be all that's required.
  • That all the images we're concerned with will be thumbnails, clicking on which will take the reader to a larger version along with the licensing information, muddies the water even further, and makes it difficult to implement any kind of longdesc solution.
  • The expert opinions so far gathered have failed to recognise this fundamental distinction between HTML alt text and the way that alt text is used in the {{tl:Image|}} template.
  • We need to be breaking new ground here, not following like sheep. HTML alt text is certainly an alternative for an image, but {{Image}} alt text should be a short and succinct description of the image. I think we were heading on the right lines the way this guideline was developing; perhaps the rest of the world will catch up with us one day.

--Malleus Fatuorum 20:49, 19 March 2010 (UTC)

Is there a difference between this {{Image}} template you mention and a naked image generated by [[Image:NAME.jpg]] or [[Image:NAME.jpg|thumb|caption]]? Would HTML alt text be like the alt text used for (as an example) the Wikipedia logo (at top left in the skin I'm using)? Plus alt text for all the buttons like "Save page"? Carcharoth (talk) 21:58, 19 March 2010 (UTC)
I'm a little puzzled by your question. How do you include images in your articles? I've got no interest in the decorative stuff like the wikipedia logo; I thought we were discussing images in articles? --Malleus Fatuorum 22:12, 19 March 2010 (UTC)
Sorry, I've just realized that I'm talking bollocks. Of course you're right, I meant the wikimedia software thing [[Image:NAME.jpg]], not the template, which makes the whole disussion moot, as getting the software changed is nigh on impossible. --Malleus Fatuorum 22:17, 19 March 2010 (UTC)
The usability people might be able to get traction. By this I mean the WMF usability people, though I don't know how much they look at accessibility stuff. There is also an en-Wikipedia accessibility group. WP:ACCESSIBILITY and Wikipedia:WikiProject Accessibility. Carcharoth (talk) 22:27, 19 March 2010 (UTC)

Archiving

There was more recent discussion than I had realised. Could those who participated in the discussions indicate which bits have finished being discussed? Maybe just archive it all after all and let people individually bring sections out if they want to continue earlier discussions? Carcharoth (talk) 21:08, 19 March 2010 (UTC)

Thoughts from a screen reader user

I agree with Jared that the content and function of the image is the most important for alt text, but, as a blind person, I like to know what an image looks like, or what the main features of an image are. I also agree with him that for some images, the caption suffices for the alt text. However, that's difficult to implement in Wikipedia, since all images have links to the image description pages, and image captions must be visible to readers. I'm not sure what the optimal solution is here, either, and I think it might involve changes to MediaWiki. Maybe the default alt text should be changed to "see caption" or similar, or perhaps, as Jared says, we should somehow "combine the image and the text caption into one link". Simetrical has dealt with alt text in the past, and, if we get consensus here for a change to the behaviour of MediaWiki, we should notify him. Graham87 06:03, 20 March 2010 (UTC)

General thoughts

There is a wide spectrum of possible lengths and styles for alt text. I would urge a re-reading of this discussion (which I just archived, ironically enough), as it is important to get the views of those that describe images as well as those who the images are being described for. I have experience with describing/captioning/explaining images for sighted people, but have little experience of doing so for blind or partially-sighted people. What would be good is to get opinions from people who professionally describe images for blind people, and when to be brief and when longer descriptions are needed and where to put those descriptions.

My view is that longer descriptions (longdesc) should go on the image file page (whether here or on Commons) along with a more detailed image-focused caption (lots of detail suited for anyone seeing the image in isolation, arriving there from a variety of locations), and a shorter, more succinct version should be used for the alt text that should be focused on the context of the article it is used in. But the critical point is to have things set up so the screen reader tells the reader that they have reached an image, and for the caption and the alt text to then work together to tell the blind reader (like any other reader) what they are being "shown", in the context of the article.

The question is whether the screenreader encounters the alt text first and then the caption, or the other way round? Think of a sighted reader, who scans the image, briefly registers what it is, then reads the caption, and then goes back to the image to reassess what they are seeing in light of what the caption said. This switching back and forth between the image and the caption is quite common. How is that done with a screenreader would be my question. Is it easy for those using screenreaders to switch back and forth? If so, then the caption and the alt text should be written together, with this in mind. If not, then the order in which the two are read needs to be considered.

The other thing to bear in mind is that there is disagreement among those surveyed as to what is needed. It is important not to cherry pick one or two opinions from a survey and use that to impose conditions (some people want longer descriptions, some want shorter ones). Also, many of the surveys focused on consumer websites where a particular service was provided (shopping, banking, telephone services, news, etc). I didn't see, in any of the surveys, examples of surveys of educational information content websites like Wikipedia. It is possible that the correct approach for a website like this hasn't been worked out yet.

For some images, full descriptions approaching those used in educational establishments like those used in art museums and galleries and those used to describe videos and plays to blind people (audio description), will be needed. For others, only brief descriptions will be needed. It all depends on how the caption has been written, and what the purpose of the image is in the article (as an aside, for featured pictures, I would suggest a very long and full description approaching that which would be used to describe a work of art). If a long caption is needed, long alt text will generally be needed (though repetition of the caption should be avoided). If the caption is short, then the alt text can probably be short as well (though there are exceptions, generally when someone choses to use a short caption and let the image speak for itself, something that has to be done using alt text description for those that can't see the image).

But whether short or long, the key is usability and comprehension. Will those using screenreaders understand what images were used and why, after reading the article, and were they able to navigate through the article without confusing notifications that there is an image here but no description of why there is an image here and (if needed) what the image is showing? If not, then something could be improved somewhere.

A good resource is here, with a page on verbal description here. That is probably beyond what alt text is designed for (and possibly beyond the volunteer resources here), but gives an idea of what could be done. There is also a bit on accessible web materials here. Maybe it is possible to view Wikipedia as a musuem with screenreaders giving audio descriptions of the exhibits (articles) and the images in the articles? Which brings me round to the spoken articles. How do those doing spoken versions of articles handle images? Do they do full audio descriptions of the images? WP:SPOKEN mentions this here. I'll drop a note there to see if any of the people active there have opinions on the discussions on this page.

Sorry for the length of the above. I'll split it up into smaller sections if that is needed, or please respond in subsections if that helps. Carcharoth (talk) 21:50, 19 March 2010 (UTC)

We seem to be just going round and round in circles. Context has been mentioned before, but not in the right context. We are talking very specifically here about what is the appropriate text to include in the alt parameter of [[Image:someImage.jpg]], given the ubiquitous presence of a caption for every image we're concerned with, not about HTML alt text. --Malleus Fatuorum 22:05, 19 March 2010 (UTC)
"It is possible that the correct approach for a website like this hasn't been worked out yet." Not only possible, self-evident. --Malleus Fatuorum 22:08, 19 March 2010 (UTC)
I'm glad we agree on that last point, as there was hostility the last time I suggested that guidelines written for general web developer might not be 100% applicable. Responding to the point Malleus made in the section above: not all images in featured material have visible captions. Featured lists often use images within tables without captions.
Wrt going round in circles, I'm ready for a break and to come back fresh. The paper posted by Carcharoth/Slp1 indicated there is in fact little agreement among experts or blind users as to what they want from alt text. We mustn't just seize on a single expert's opinion just because it is all we have. The needs, education and patience of blind users is likely to vary just as much as sighted ones. Some of our readers will read a huge FA from beggining to end, but most won't go further than the lead. Colin°Talk 23:07, 19 March 2010 (UTC)
I suspect we may agree more than we disagree. So far as images in featured lists are concerned though, they're always described in an accompanying column, so the caption is just in a different place if you like. --Malleus Fatuorum 23:21, 19 March 2010 (UTC)
In both JAWS and Window-Eyes, the most popular commercial Windows screen readers, the alt text is read before the caption. This is also true of NVDA, a free Windows screen reader.
Re: Alt text length, anything longer than a few sentences should go on the image description page. If this is done, the alt text should mention that more information can be found by "activating this link" (don't use "click here!). Graham87 04:52, 20 March 2010 (UTC)
Gah, I didn't notice the section above about external reviews, since I tend to read discussions backwards (i.e. the newest heading first). I'll have to think about this some more. Graham87 05:06, 20 March 2010 (UTC)
Your opinion is the one I'm most interested in Graham; the rest of us are just making educated guesses at best. --Malleus Fatuorum 05:26, 20 March 2010 (UTC)
Hi Graham. Thanks for commenting here and answering the question about what order the alt text and caption is read out in. I guess questions like that can be answered by setting up a screenreader and playing with the settings, but hearing it from a screenreader user is great. The point you make about preferences or options is also interesting. Do you think we should assume default settings are used, or assume some obvious preferences are enabled? I read somewhere that it is fairly easy to skip things once you are used to using a screenreader, and focus on what you want to read, but I'm less certain about returning to previous points in a text using a screenreader (i.e. going back to earlier parts of the text). Is it possible when using a screenreader to skip back and forth between adjacent text like alt text and a caption if needed? Carcharoth (talk) 10:22, 20 March 2010 (UTC)
Yep, moving backwards through text is as easy as moving forwards through text. Also, it's best to assume that default settings are used. Graham87 11:29, 20 March 2010 (UTC)

Non-specialist knowledge

Where is the threshold of "non-specialist knowledge"? I think it needs to be somewhere above "something everybody knows" otherwise we're left with the lame "A man", "A boat", "A church". But it will be below something that only an expert in the subject could know. Here's some examples:

  • A person's hairstyle
  • An unusual item of clothing (e.g. historical or occupation-related)
  • A type of boat
  • The parts of a church
  • A type of tree such as an oak or beech.
  • A fashion

Some of the sources talk of blind readers feeling they are missing out. Clearly we can't fully replace the image, but our seven words or so should use better adjectives than those available to a four-year-old. Thoughts? Colin°Talk 10:37, 20 March 2010 (UTC)

BTW: I'm finding this talk page very difficult to navigate on my netbook. I'm happy for much of the pre-external-review stuff to go. Colin°Talk 10:37, 20 March 2010 (UTC)

Done. SlimVirgin TALK contribs 10:55, 20 March 2010 (UTC)
I've unarchived the Fangs section, since I'd made a recent comment there. Graham87 11:27, 20 March 2010 (UTC)
For me, "something everybody knows" equals material that won't be challenged. If you describe a hairstyle in an image as Buzz cut and I describe it as Crew cut then it's being challenged.. Expert opinions are only useful if they come with WP:RS about an image. Regards, SunCreator (talk) 18:30, 20 March 2010 (UTC)
I agree. We need to keep it simple so that anyone looking at the image would agree with the description, if it's a description we want. Jared Smith was suggesting function mattered more than a description of the image's content, so I hope we'll be hearing more about that distinction. SlimVirgin TALK contribs 18:35, 20 March 2010 (UTC)
To me, function is "this image is being used as the lead image to show the appearance of the subject of this biography", or "this image is being used to show the subject of the biography at a younger/older age", or "this image is being used to show the aftermath of this natural disaster", or "this diagram is being used to show the reader parts of a bird's anatomy", while a description is describing what is shown (i.e. type of objects [human, animal, building], type of picture [panorama, landscape, portrait, micrograph, x-ray, telescope image, satellite image], and some details of colour, size and appearance where necessary). Some of the description will have been covered in the caption, some won't have been. The function of the image depends on the article it appears in (in one article it might simply be part of a list or gallery, or used in a section of an article, in another it might be the lead image). But the description of the appearance of an image is largely independent of the article it appears in, so that could be left to the file page, with only the most basic bits coming through to the article. It is important when writing alt text to think "does it apply to the image wherever it is used, or just to this article?" If it applies to the image where-ever used, then it is better put on the image page. If it applies only to the use of the image in this article, then it should be in the caption or some alt text attribute that is specific to the page, not to the image. Is that making sense? Carcharoth (talk) 19:45, 20 March 2010 (UTC)

I don't really want this section to be a debate on function vs descriptive and decorative vs informative images. Those issues are important but let's keep focussed on this one. If we are describing the image, albeit briefly, what is the threshold for "non specialist"? The "challenged or likely to be challenged" test doens't apply. We have a source, a primary sources, that the original and challenging editor can both consult. The issue is describing that source, which is the same as for a book, a journal article or a film source. Let's say the picture of a haircut appears in our article on haircuts. Someone uploads a picture of themselves in a crew cut and makes that the caption. Surely "Crew cut" as a caption is just the same as "Boy with a crew cut" as alt text. Unlike some famous painting or photo, there's no published commentary on the image to prove it is a crew cut. But editors can view it and if they disagree then come up with an acceptable alternative, etc. My preference is set the level at "knowledgable about the subject, but not a specialist in the subject". That way, we don't let the general-ignorance limit us to just "a boy". Anyone who knows anything about haircuts will recognise a crewcut. Anyone who knows anything about boats will recognise that boat as a yatch. Someone who knows nothing about XXX might struggle. And so on. Colin°Talk 20:27, 20 March 2010 (UTC)

External reviews

I think we need to have a major rethink of this guideline based on this external review I have received from Jared Smith from webAIM (www.webaim.org). Especially as my other external review correspondent indicated that he would also make very different recommendations from those currently given- without giving details.

"I'm not sure where to start. Their presentation of and approach to alternative text is fundamentally flawed. This gross misinterpretation begins with the first sentence - "Alternative text (alt text) is a description of an image that text-only users will see instead of the image." This is *NOT* at all what alternative text is! Alternative text is an alternative to the image. It is NOT a description of an image. What an image looks like is almost universally quite different than what the content of an image is. The wikipedia approach is also quite contrary to the HTML and WCAG specification's definition and recommendations for alternative text.

Nearly all of the examples posted get this wrong. Someone that is blind or that has images disabled doesn't care what an image looks like, they want the *content* and *function* of the image. In many of the examples the alternative text for the image should be identical to the caption - and because the caption is already presented, the images should thus be given empty alternative text except in cases where the image is linked in isolation, in which case redundant or very simple alternative text is required (an image that is the only thing within a link MUST have alternative text).

For example, in the Van Gogh example, the full content of the image is already presented in the caption. Someone that is blind does not care about his hair style and a description of his hair style is not an alternative to the content of the image. It's Van Gogh, that's all that matters. There's fundamentally and technically no difference between their flawed alt text and their recommended alt text, except that one happens to be shorter in length. In most cases, this image would be given empty alternative text (alt=""), but because the image is linked, it must be given alternative text - probably alt="Vincent van Gogh" or similar. A more optimal approach would be to combine the image and the text caption into one link with the image being given empty alternative text.

I would attempt to edit this document, but to make it accurate and most optimal for end user accessibility would require significant changes to what seems to be a well-accepted, though highly misunderstood and flawed approach to alternative text at wikipedia. But, you are welcome to post my comments and point them to http://www.webaim.org/techniques/alttext/ for further guidance."

I have responded that the guideline is undergoing major changes at present, and his comments likely to spark more. I also assured him that any direct input and edits he could make would be welcomed, and encouraged him to participate if he had the time. --Slp1 (talk) 13:09, 19 March 2010 (UTC)
This is pretty damning. While some of our sources do support the approach we have taken, this is the most authoritative external and expert opinion we have. The chap is a web accessibility expert (as opposed to merely a web designer) and has considered Wikipedia's needs rather than, say, a commercial website. My only consolation is that the mighty Matt Cutts of Google is clearly barking up the wrong tree too :-)
I propose this guideline be demoted immediately, some sort of "Caution: this might be bollocks" template be applied to it, and it be removed from the FL and FA criteria. I feel bad for Eubulides and SlimVirgin who have put an enormous amount of work into it. I still strongly believe WP should have a "Alternative text" guideline, and that that guideline should be reflected by "our best work", but it isn't this one.
I'll let you know if I hear anything from the experts I've contacted. Regards, Colin°Talk 13:57, 19 March 2010 (UTC)
I think the comments of your expert are pretty much what many of us have been saying for some time now, and pretty much what Matt Cutts was saying as well. The only substantial difference here is that almost all of the images we're concerned with in articles will already have a caption that may suffice, meaning that alt text isn't required, or only simple (redundant) alt text because of the linking issue . Agree with Colin that this guideline needs to be demoted and flagged now, and the requirement for alt text removed from the FA criteria. --Malleus Fatuorum 14:11, 19 March 2010 (UTC)
I must have missed the bit in Jared Smith's advice that the purpose of alt text was to help Google index our pictures ;-). Colin°Talk 14:49, 19 March 2010 (UTC)
I support the guideline status being removed until we get this clarified, and in the meantime it really ought to be removed from the FA criteria, because we may actually be making things worse for people using screen readers. SlimVirgin TALK contribs 14:23, 19 March 2010 (UTC)
I agree that a radical rethink and some major edits are likely required. However, I, with Colin, strongly believe in principle that Featured Articles, our best work, should reflect the best practices for web-accessibility, which includes alt-text. That the current guideline doesn't reflect best practices doesn't mean that we should abandon the principle. Already significant changes have been made to this guideline to address concerns thanks editors here: more can and obviously should be made. Before making any major decisions, let's give ourselves a week or so to see if the problems can be addressed to everybody's satisfaction, and to get additional outside reviews and help if possible. Once again, I repeat my offer to change the alt-text of recent Featured articles so that it conforms to the best practice guideline that eventually emerges. --Slp1 (talk) 14:39, 19 March 2010 (UTC)
That's very generous of you Slp1, but it doesn't make up for the time spent writing overlong text and getting it reviewed. Plus, it really will be hard to muster enthusiasm for writing against a contested guideline or one where the text changes every time you look at it. I did offer a while back for this guideline to be pruned hard to something we can agree on. Perhaps that should be the next step, but I think that can be done without it being a guideline. Unstable guidelines aren't helpful. Colin°Talk 14:49, 19 March 2010 (UTC)
Sp1, Jared Smith made the point that the caption will often suffice as alt text (with the parameter filled in so that it's not read out twice). This is something I've been asking about for a while, including on this page, and for some reason there was resistance to it, though I can't understand why. It means that the advice always to add extra words in the alt text is flawed, and yet that's what we're requiring all new FACs and all FARs to do, though it can take considerable work in articles with lots of images. That criterion needs to be removed pretty quickly. SlimVirgin TALK contribs 15:10, 19 March 2010 (UTC)
I find it very confusing when people say "caption will often suffice as alt text" (which is clarified in SV's text) or that "alt text isn't required". What should a reader think we mean by "alt text"? There's "text that is an alternative to seeing the image", which may be satisfied by the visible caption; there's "text given for the alt= parameter in wiki markup" and "text that appears in the alt= parameter in the HTML". This confusion has certainly led to my rejection of proposals that "we use the caption for alt text" as I interpret it as option 2 (the alt= wiki markup). What we are really saying is that "the caption will often suffice and the alt text in the HTML should be blank". But it seems that achieving blank HTML alt= text isn't always possible. In addition, I find it very confusing when this guideline talks about "caption" when it means the caption field in the image wiki markup and the example has no visible caption. Perhaps we need to agree on our terminology. That's something for another section I suppose... Colin°Talk 15:44, 19 March 2010 (UTC)
That is pretty rough from the expert. So with policy concerns addressed, our basic principle behind this guideline is flawed. In light of that, I think this whole page is so far off we might as well archive it and start from scratch. In regards to FAC, I see no problem with requiring some basic alt text solution (if that is defaulting to the caption, that's fine) but this guideline has got to go. —Charles Edward (Talk | Contribs) 15:20, 19 March 2010 (UTC)
Whaw, didn't think any 'expert' would put it so clearly. I am in broad agreement with Jared Smith's response, of 'In many of the examples the alternative text for the image should be identical to the caption - and because the caption is already presented, the images should thus be given empty alternative text except in cases where the image is linked in isolation, in which case redundant or very simple alternative text is required' and 'Context is Everything'. His link is also excellent. Regards, SunCreator (talk) 15:48, 19 March 2010 (UTC)


This talk page is getting pretty long. Given the above comments and likely change of direction, pretty much most of the above sections are unlikely to be worth pursuing with further discussion. Time to archive? Colin°Talk 15:51, 19 March 2010 (UTC)

Yes, archive now seems good timing. Regards, SunCreator (talk) 16:04, 19 March 2010 (UTC)
(ec) I think we have no choice but to demote it given the expert response. All we're able to do ourselves without specialist knowledge is copy edit it, but if the whole approach is flawed, more copy edits aren't going to help. SlimVirgin TALK contribs 15:53, 19 March 2010 (UTC)
What we need now is for someone to tell us what to write to make the caption the alt text so we can advise at FAC. Jared Smith says alt="", but I was told on this page that was wrong. Should we be writing alt=|link= or something else? SlimVirgin TALK contribs 15:55, 19 March 2010 (UTC)
Colin, to address your concern above, the problem seems to be that we were told every image must have alt text, when in fact it's that every image must have an alt attribute, if you read Jared Smith's link. More text is often not needed because the caption suffices. All we need to do is make sure the screen reader doesn't read it twice or doesn't read the file name or something else. SlimVirgin TALK contribs 15:58, 19 March 2010 (UTC)
Agree, we are singing from the same hymnsheet now but still tripping up on terminology. Could I suggest we don't use the abbreviation "alt" unless we mean the wiki markup. And probably avoid referring to the HTML markup unless we are having a technical discusion. Colin°Talk 16:03, 19 March 2010 (UTC)
SlimVirgin, I suggest demoting it for a set period of time. Two weeks or a month perhaps. That should give it time to be revised completely before it comes back. Regards, SunCreator (talk) 16:14, 19 March 2010 (UTC)
I've removed the guideline tag. I didn't replace it with anything. What I suggest we do is spend some time trying to figure out what a guideline should say, rewrite it, add the proposal tag, then ask around for fresh eyes and consensus. It was never proposed for guideline status anyway so this is somewhat overdue. Hopefully we can come up with something that works for everyone. SlimVirgin TALK contribs 16:38, 19 March 2010 (UTC)

SunCreator, I've received an email from Carcharoth, who can't post until he gets home, but he asks that you unarchive some of the recent posts, including one from him that he would like people to be able to read. He also passed on this article about alt text from City University in London, which looks interesting. SlimVirgin TALK contribs 17:02, 19 March 2010 (UTC)

Thanks Carcharoth, the article looks like an excellent resource; very much to the point, and somewhat different recommendations from Jared Smith. I'm going to do some journal database searches to see about other sources. I should have thought of that before. --Slp1 (talk) 17:14, 19 March 2010 (UTC)
Apologies to Carcharoth. Can someone unarchive for me. I'm out to eat and can't cut/paste so much from my mobile. Will have access to a PC in about an hour and will correct that if it's not done before. Regards, SunCreator (talk) 17:21, 19 March 2010 (UTC)
I've unarchived the whole thing. Perhaps we can rearchive some of it once we're agreed on what needs to stay. SlimVirgin TALK contribs 17:26, 19 March 2010 (UTC)
Thanks for passing on the message. It wasn't necessary to archive or unarchive the whole page. I just wanted to be sure (as someone who arrived late to the discussion) that my comments here and here were not prematurely archived. What I will do now is go back up the page and archive sections that are more than a week old. And then someone could set up an archive bot maybe? I don't really know how to do that. If more-recent discussions need to be archived to start again with a clean sheet, fine, but the key here is, I think, to realise how image captions and alt text need to work together. Carcharoth (talk) 20:45, 19 March 2010 (UTC) There is an archive bot set up already. algo = old(21d) - does that mean every 21 days? Will the bot archive to archive 3 or the newly-created archive 4? Will manual archiving mess things up?

Jared Smith has sent an updated email last night. Since I haven't got explicit permission to post it, I'll just paraphrase. He's happy that the guideline is being reconsidered and hopes he hasn't been "overly critical". He writes that alternative text is difficult to understand, and even more difficult to define and write specifications for. He feels the current guideline has the potential, with some work, to be "a wonderful resource". He would like to help, but is very busy for the next few weeks. In the meantime, he is willing to answer any specific questions we have. Maybe we can collect any queries for him, and I will act as liaison in emailing him. --Slp1 (talk) 13:40, 20 March 2010 (UTC)

Please tell him we're taking the criticism constructively, and it means we're all reading about alt text, some of us for the first time. :) SlimVirgin TALK contribs 14:12, 20 March 2010 (UTC)
I am thinking of writing an email to Jared in a few hours with a couple of questions. In particular, I would be interested in getting his feedback on the recommendations etc from the Peltrie article. I am also curious about what he thinks about whether (and how) WP's status as an educational resource should affect our alternative text: I am intrigued by Carcharoth's suggestion that we think about audio descriptions at museums as possible models for longer text descriptions. I've been really enjoying this BBC/British Museum A History of the World in 100 Objects podcast which attempts much the same thing for radio listeners.
Does anybody have other questions that they would like me to ask?--Slp1 (talk) 14:27, 21 March 2010 (UTC)
  1. ^ Lazar J, Allen A, Kleinman J, Malarkey C (2007). "What frustrates screen reader users on the web: a study of 100 blind users" (PDF). Int J Hum Comput Interact. 22 (3): 247–69. doi:10.1080/10447310709336964.{{cite journal}}: CS1 maint: multiple names: authors list (link)
  2. ^ Buzzi M, Leporini B (2008). "Is Wikipedia usable for the blind?". Proc 2008 W4A. Beijing. pp. 15–22. doi:10.1145/1368044.1368049. {{cite conference}}: Unknown parameter |booktitle= ignored (|book-title= suggested) (help)