Wikipedia talk:Alternate text for images/Examples

I have done a survey of Wikipedia images, using Special:Randompage until I'd found 100 instances of uploaded images in Wikipedia articles, and writing the best alternate text I could for each.

The results

adenine.jpg in Adenine: "Adenine is a chemical with the empirical formula C5 H5 N5." (not an exact equivalent, but a full textual explanation of its structure would be disproportionately long in such a short article)
Hexagon.png in Hexagon: "" (meaningful, but meaning already in text)
- agreed
BadfingerNoDice.jpg in No Dice: ""
- disagreed - "the cover of the album shows (etc)" is appropriate here.
Caligula.jpg in Caligula: ""
Ap13.jpg in Apollo 13: "The Apollo 13 logo featured three flying horses, and the motto 'Ex luna, scientia'." (this wouldn't normally be important enough for such a short article, but it's ok since it's part of a list of miscellaneous info about the mission)
Zinc_finger.png in Zinc finger: "" (meaningful, but meaning already in text)
Chirac2.jpg in Jacques Chirac: ""
- disagreed: "Chirac has short, receding black hair, brown eyes, and often wears a suit."
- Find me any other encyclopedia which considers Chirac's receding hair or brown eyes important enough to write about, in an article up to five times as long as Wikipedia's (just to make it really easy!), then I'll take that suggestion seriously. -- Mpt
  - EPNWorld Reporter. It's a current events website, not a dedicated encyclopedia, but the point stands. The appearance of world leaders is a subject of considerable interest, and rightly something that should be covered in Wikipedia. Using images for those who can access them, and a text alternative for those who can't.
  - If you mean the Schröder lawsuit, that might indeed be sufficiently interesting to include in his article, but it would almost certainly be non-alternate text -- because I doubt you could find any image that would successfully convey the necessary details about the lawsuit. And the rest of Schröder's appearance still wouldn't be important enough to write about. -- Mpt
  - You think it isn't important enough. I think it is important enough. Our opinions differ.
Cumulonimbus_thumbnail.jpg in Cumulonimbus: "" (appearance already in text)
- disagreed: "the clouds can appear striking..." ["... during sunset."]
- When the article is long enough to have a section on "How to photograph cumulonimbus" without looking silly (about ten times as long as it is now, maybe), then you might have a point. Until then, why is how they look during sunset any more important for the text of an encyclopedia article than how they look during sunrise or at midday? (Hint: "Because I happened to have an image of it in the visual rendering" is not a valid answer.) -- Mpt
  - Their appearance at mid-day is covered in the main text (see references to anvils).
  - So why not move the image to that part of the text, so the article doesn't sound jumbled -- e.g. "During the spring and summer, cumulonimbus clouds are more likely to form in the afternoon, due to the heating of the earth’s surface[[Image:...|, but they look even more striking at sunset]]."? Because, as with my dilithium-3 example, that's not what the image is intended to convey. It's intended to convey what cumulonimbus looks like in general, and that's already in the non-alternate text. (Hmm, perhaps it shouldn't be.) -- Mpt
  - No - that para is about cumulonimbus formation. The second para is the one that is (partly) about their appearance. Both image and alternate text are thus well-placed, though they could be improved if/when the Wikipedia software is changed to allow mid-paragraph floating images. The purpose of the image is to convey BOTH what the clouds look like in general AND what these specific examples of the clouds look like. The former is in the text. The latter is, correctly, in the alt text.
Ngc6240_composite.jpg in NGC 6240: ""
Cray.jpeg in Cray Research: ""
TheBandMusicFromBigPink.gif in Music From Big Pink: ""
- disagreed: Again, "the cover of the album shows..."
Nero_wolfe_stamp.jpg in Nero Wolfe: "Wolfe was featured on a postage stamp issued by San Marino in 1979."
DixieHighwayMarker.JPG in United States highway: "" (appearance already in text)
JeffersonHighwayMarker.jpg in United States highway: "" (ditto)
State_shield.gif in United States highway: "" (ditto)
Five_forks.jpg in Battle of Five Forks: ""
Nas.jpg in Nas: ""
GZALiquidSwords.jpg in Liquid Swords: ""
C_str.jpg in Candystriper: "" (appearance already in text)
Piusv.jpg in Pope Pius V: ""
John_dryden.jpg in John Dryden: ""
StopEssoleaflet.jpg in Stop Esso campaign: "A leaflet produced by the campaign, for example, said 'Don't buy E$$o' with the parody logo, together with a gloved hand setting fire to the Earth." (I've updated this one myself)
Joyful3.jpg in Rosary: ""
Sorrow5crop.jpg in Rosary: ""
Glorycrop.jpg in Rosary: ""
Glory5crop.jpg in Rosary: ""
StevieWonderSongsintheKeyofLife.jpg in Songs in the Key of Life: ""
Ei-map.jpg in Geography of the Republic of Ireland: "" (location already in text)
Www.wesleyjohnston.com-users-ireland-maps-island_agriculture.gif in Geography of the Republic of Ireland: "Agriculture in the north of the Republic is dominated by small dairy and cattle farms. Dairying also features in the southwest, along with arable farming in northern and eastern Munster. Grazing is predominant in central Ireland, with livestock and arable farming in the southeast. Dublin and Wicklow have mixed styles of agriculture. Coastal areas in the west are only sparsely cultivated."
Asia.jpg in Asia: "" (location etc. already in text)
Husum.jpg in Husum: "The town's official seal features two blue lions in a wooden fence's roofed gateway, with three red flags flying from the roof." (If this was used rather than "", however, the image itself would need to be placed much later in the article, as Husum's seal isn't as important as Theodor Storm is.)
Ukraine_flag_medium.png in Ukraine: "The nation's flag consists of two equal-sized horizontal stripes -- the top one light blue, the bottom one yellow."
Ukraine_coa.png in Ukraine: "The Ukrainian coat of arms features these colors too: a blue shield with yellow trident, supported by a crowned lion on the left and a figure in traditional dress on the right."
Rio_cover.jpg in Rio: ""
The_swan.JPG in Grace Kelly: ""
- Disagree: [[Image:the_swan.JPG|In the 1956 film The Swan, Kelly starred in the role of a princess, alongside Alec Guinness and Louis Jourdan.]] is better than nothing. The film is not even mentioned in the article, other than in the filmography, so it can hardly be said that the image repeats information already present.
- That's right, it doesn't, because it's not important enough for the article text, other than the filmography. Why are you singling out The Swan for special attention in the text-only rendering, and not any of the other movies? (Hint: "Because I happened to have an image of it in the visual rendering" is not a valid answer.)
D%27AngeloVoodoo.jpg in Voodoo (album): ""
Vineyardmap.png in Martha's Vineyard: "" (location already in text)
Br_PE.png in Pernambuco: "The state's flag has sections of dark blue for the sky and white for the land. In the sky section is a star, a rainbow, and a smiling sun; on land is a red Christian cross."
Nicole_Kidman.jpg in Nicole Kidman: ""
Expresso.jpg in Espresso: ""
- Disagree: [[Image:Espresso.jpg|Espresso appears dark brown with a gold-colored foam on top, and is often served in small portions.]] repeats information already in the article, but is a good summary of what espresso looks like. (Image has been renamed, by the way)
- That's your best attempt yet. But the first half is apparently wrong (it's only straight espresso that's dark brown) and redundant (surely any straight coffee is dark brown!), and the second half is also redundant ("is often served in small portions ... is served in small amounts called shots"). A listener, hearing this repetition, would growl at your sloppy copy-editing. -- Mpt
US-DeptOfTransportation-Seal.png in United States Department of Transportation: "The Department's seal consists of three blue droplets following each other in a circle, with the words 'Department of Transportation' and 'United States of America' arranged around the edge of the circle."
Thompson.jpg in Thompson submachine gun: ""
- Disagree: [[image:thompson.jpg|The Tommy Gun is often pictured with one of its most characteristic features, the 100-round drum ammunition cartridge.]] might be a good start. The gun's appearance is barely mentioned in the article.
- Hey, you're improving. :-) But I can't tell how many rounds that cartridge is by looking at the picture, so if it's worth mentioning in the article at all, it should be outside the alternate text for the sake of us poor fully-sighted people. And discussion of how "The Tommy Gun is often pictured" would make sense in a section on "Tommy Guns in entertainment" or "Tommy Guns in art", but the article has neither. Net result: zero alternate text. -- Mpt
Glamis_castle.jpg in Glamis Castle: ""
ArthurBrownTheCrazyWorldofArthurBrown.jpg in The Crazy World of Arthur Brown: ""
JEdgarHoover.jpg in J. Edgar Hoover: ""
US-DeptOfVeteransAffairs-Seal.png in United States Department of Veterans Affairs: "The Department's seal features an eagle clutching the flags of the USA and ______, and five gold stars above; with the words 'Department of Veterans Affairs' and 'United States of America', together with a gold braid, arranged around the edge of the circle."
Heinlein_book.JPG in The Man Who Sold the Moon (short story collection): ""
JohnAM.jpg in John A. Macdonald: ""
ATP.png in Adenosine triphosphate: "" (structure already explained in text)
Ricefield.jpg in Rice: ""
KentCountyMD.jpg in Kent County, Maryland: ""
PaulTibbets.jpg in Paul Tibbets: "" (Enola Gay already explained in text)
Dio_coin1.jpg in Diocletian: ""
- Disagree: [[Image:Dio_coin1.jpg|Diocletian's likeness is featured on coinage along with the phrase IMP CC VAL DIOCLETIANVS PF AVG.]] tells us what is on the coin pictured, and is not mentioned elsewhere.
- Diocletian was surely featured on many things more magnificent than coins, like official portraits or statues. Why are you singling out a coin for special attention in the text-only rendering of the article? (Hint: "Because I happened to have an image of it in the visual rendering" is not a valid answer.)
Piaf1.jpg in Edith Piaf: ""
Piafgrave.jpg in Edith Piaf: ""
Acetal.jpg in Acetal: "R-C-O-R', with the C also having an -O-R branch,"
Shrine_of_bahaullah.jpeg in Bahaullah: ""
William_Alexander_Henry.jpg in William Alexander Henry: ""
SMarceau.JPG in Sophie Marceau: ""
JackPickford.jpg in Jack Pickford: ""
California_map_showing_Yolo_County.png in Yolo County, California: "" (location already in text)
0771aa.jpg in The Humility of Pain: ""
ElectoralCollege1789.png in U.S. presidential election, 1789: "There were 69 voting electors: ten from Massachusetts, five from New Hampshire, seven from Connecticut, ten from Philadelphia, six from New Jersey, three from Delaware, six from Maryland, ten from Virginia, seven from South Carolina, and five from Georgia." (how they voted, and why some didn't, is in the immediately following text)
Us-ia.gif in Iowa: "The state's flag features a bird holding a banner aloft on a white field, which is bordered by a blue stripe on the left and a red stripe on the right."
Iowastateseal.jpg in Iowa: "The state seal includes a similar bird and ribbon, as well as a settler holding a United States flag; the words 'The Great Seal of the State of Iowa' are arranged on a yellow ring around the outside."
US_ia_highlighted_map.jpg in Iowa: "" (location already in text)
Tourmaline.xtal.250pix.jpg in Tourmaline: "" (appearance already in text)
Boscovich.jpg in Rudjer Josip Boscovich: ""
Nectarinetree.jpg in Nectarine: "Nectarine blossoms are typically pink in color."
Isabella_of_Castile.jpg in Isabella of Castile: ""
Lice_image01.jpg in Lice: ""
- Disagree: [[image:Lice_image01.jpg|Sheep lice have translucent bodies and a proportionally large head and abdomen.]] (may not be the best replacement, but I think it's better than nothing)
- Here the lice are in the same boat as poor Grace Kelly. Why are you singling out sheep lice, of all the lice species, to describe them in the text-only rendering of an article about lice in general? (Hint: "Because I happened to have an image of them in the visual rendering" is not a valid answer. And maybe you could clarify what "a proportionally large head and abdomen" means, too.) -- Mpt
Rhineland_rheinhessen.png in Rheinhessen-Pfalz: ""
Tetris-gravity.png in Tetris: "" (a text equivalent of the example would be much harder to understand than the explanation already in the text)
Iberia.a320-214.ec-hdl.250pix.jpg in Iberia Airlines: "" (description of airline insignia not important enough for such a short article)
- Disagree: [[image:iberia.a320-214.ec-hdl.250pix.jpg|Iberia's insignia consists of the stylized letters IB in red and orange, with colored stripes running the length of the aircraft.]] serves as a replacement, since it conveys similar information to that conveyed by the image.
- Yes it does, but it's in the same predicament as Monsieur Chirac's eyes and hair: the article would need to be considerably longer before the insignia would be worth mentioning. Even when that happened, it would likely happen outside the alternate text, because the same sentence would mention when the insignia was adopted and what it represents -- things which an image itself can't do for sighted readers. (And I can tell you haven't listened to how funny the start of that article sounds ...) -- Mpt
Th-map.png in Geography of Thailand: "" (location already in text)
William_Jennings_Bryan.jpg in William Jennings Bryan: ""
ElectoralCollege1900.png in U.S. presidential election, 1900: "The electoral college was split almost completely North-South, with northern states electing the Republican (McKinley) and southern states the Democrat (Bryan). The exceptions were the Northern states of Montana, Idaho, Nevada, Colorado, and Missouri, which also supported Bryan. But McKinley won with 65 percent of the electoral vote, versus 35 percent for Bryan; the popular vote also went McKinley's way, 52 percent to 46 percent."
Gogmagog.png in Gog: ""
Blekinge.png in Blekinge: "The province's coat of arms features a five-pointed crown with a navy cushion; and a stylized blue shield on which is placed a yellow plant, itself overlaid with three smaller crowns."
Svpmap_blekinge.png in Blekinge: "" (location already in text)
- Disagree: [[Image:svpmap_blekinge.png|Blekinge is located at the tip of Sweden, bordered on the North by Småland, and on the West by Skåne.]] gives directional details that are not present in the text.
- Same problem as the Tommy Gun. That's useful information, but Blekinge's neighbors aren't named in the map. Unless the map is fixed, those names should be outside the alternate text, not inside it. -- Mpt
Malta_flag_large.png in Flag of Malta: "The national flag of Malta has a white left half, with a cross in its corner, and a red right half." (rest of appearance already in text)
David_Suchet_is_Hercule_Poirot.jpg in Hercule Poirot: "" (already explained in text)
Ustinov_is_Poirot.jpg in Hercule Poirot: "" (already explained in text)
Pickerel.jpg in Pickerel: "" (appearance already in text)
Moulins.JPG in Maurice Utrillo: ""
Test_tube_true_C.png in Test tube: "" (appearance already in text)
Blue_jay.jpg in Blue Jay: "" (appearance already in text)
Sabawi_ibrahim.jpg in [[Sabawi Ibrahim al-Tikriti]: ""
Mark_twain2.JPG in Huckleberry Finn: ""
Wang_tiles.png in Wang tile: "" (example can't practically be explained in text -- other text also needs revising so as to avoid referring to the image explicitly)
Lula_da_Silva_(17).jpg in Fernando Henrique Cardoso: ""
1-over-x.png in Asymptote: "" (appearance already in text)
- Disagree: [[image:1-over-x.png|In the graph of 1/x, the x and y axes are the asymptotes.]]
- Completely redundant. The article text already discusses in great detail "the function f(x) = 1 / x ... a vertical asymptote (such as in the preceding example's y = 0) ... a horizontal asymptote (such as in the preceding example's x = 0)". A listener would growl at your sloppy copy-editing. ... Again.
1-over-x-plus-x.png in Asymptote: "" (appearance already in text)
- Disagree: [[image:1-over-x-plus-x.png|In the graph of x+x^-1, the y-axis and the line x=y are both asymptotes.]]
- ... And again!! Sheesh, these Wikipedia authors, they don't even read what they write, do they?
Possum7439.jpg in Virginia Opossum: ""
Hannover_coa.jpg in Hanover: "The city's coat of arms is an orange shield, on which is a white castle with two turrets. Between the turrets is a lion, and under the gate is a smaller yellow shield." (not the most important thing about Hanover, so the image should be further down in the article)
Cirles_of_confusion_lens_diagram.png in Circle of confusion: "" (already explained well in text -- sentence beginning "The following diagram..." should be adapted to the image's title)
Thymidine.jpg in Thymidine: "Its empirical formula is C H6 N2 O5." (not an exact equivalent, but general structure is already in text)
Zola.jpg in Emile Zola: ""
Al-map.jpg in Geography of Albania: "Major cities include the capital Tirana, as well as Shkoder, Shengjin, Durrës, Elbasan, Korçe, Vlorë, and Sarandë." (rest of location etc already in text)
Prohibition.jpg in Prohibition: "" (it might make sense to mention the practice of disposing of alcohol, but only if the article had a section on enforcement practices, which it doesn't).

What we can learn

For more than three quarters of these images (77/100), the most appropriate alternate text is nothing at all. So I was quite correct when I wrote "The best alternate text is usually nothing". I really hope, now, that people will stop changing "usually" to the vaguer "often".

Even with Wikipedia's current limitations, appropriate alternate text is not always a complete sentence.
- If the United States highway sign examples were not already explained in the text, appropriate markup would have been: <div>[[Image: DixieHighwayMarker.JPG|The white letters 'DH' on a red background, for example, indicated a]]</div> Dixie Highway. <div>[[http://www.wikipedia.org/wiki/Image:JeffersonHighwayMarker.jpg|And a 'JH' symbol, on a white stripe against a blue background, denoted a]]</div> Jefferson Highway.
- In Acetal, the image correctly occurs in the middle of a sentence. If the image is displayed, the sentence is short enough not to require extra punctuation. Only if the alternate text is used instead does the sentence become long enough to need a comma. Therefore, the comma belongs at the end of the alternate text itself.

-- Mpt 01:42 15 Jun 2003 (UTC)

Images are often used as a visual aid, and thus tend to repeat information that is already stated in the text. As a consequence, the appropriate alt text is likely to contain repetition as well. I see no reason to believe that this repetition is a bad thing, though. With a little bit of cleverness, it's easy to avoid redundant-sounding text.

It has indeed always been acceptable for the various media in a mixed-media article to repeat each other partially or completely (e.g. subtitles); but that's a red herring, since a text-only rendering is by definition single-media. If an encyclopedia article's text repeats itself, the article is obviously flawed; and whether that faulty repetition occurred in alternate text or non-alternate text is something text-only readers are (and should be) completely oblivious to. -- Mpt

In some cases, the alt text for the image can convey additional information that might be inappropriate in the text of the article. For example, the text of the Blekinge article states that Blekinge is bordered by Småland and Skåne, but the image also conveys the additional information, such as where Blekinge is in relation to these two provinces (namely, that it is South of Småland and East of Skåne). This information could also be stated in the text of the article, but it's the kind of thing that is best conveyed by an image. So it makes sense for that image's alt text to convey this information as well.

Similarly, the article on Iberia Airlines contains a photo of one of the planes in their fleet. It seems that this photo primarily conveys information about the airlines' insignia and color scheme -- information which is not stated elsewhere in the article, probably because it's the kind of thing that is best shown with an image. So again, the alt text should convey similar information regarding the insignia and color scheme.

-- Wapcaplet 15:22 15 Jun 2003 (UTC)

With all of the images for which I'm finding (IMHO) good alt text to replace the empty string, I'm wondering if "sometimes" might be a better word than either "often" or "usually" in regards to when the empty string is best. -- Wapcaplet 16:31 15 Jun 2003 (UTC)

Alternate text is a contiguous part of the text of the article. If information is "inappropriate in the text of the article", it's inappropriate in the alternate text or the non-alternate text. If it's not important enough to be in the text, it shouldn't be in the text.-- Mpt

Album covers

The only examples where I am uncertain as to my choice are the articles on music albums, which contain an image of the album cover (about ten of the above examples). It may be somewhat interesting to describe the album cover design (not to be confused with describing the image of the album cover, e.g. "Photo of album cover"). But I don't think a description of the cover is really interesting enough for an encyclopedia article, unless some element of the cover design was particularly noteworthy or controversial (in which case this is likely to be mentioned by non-alternate text anyway). -- Mpt

I think that if I'm interested in some album, one of the things I might want to know about it is what the cover looks like. It tells me a little about the style of the band, their target audience, perhaps their influences. That's why we have album covers in articles on albums - and for the same reason we should add alternate text. Martin

I challenge you to find any encyclopedia, anywhere in the world, music-specific or general, on the Web or off, which -- in the entries for albums -- describes covers textually when they're not particularly noteworthy or controversial. (An example of "noteworthy" would be the celebrities on Sgt. Pepper's Lonely Hearts Club Band, but in Wikipedia that's already dealt with in great detail in the non-alternate text, so the alternate text would still be "". An example of "controversial" would be any case where the cover was censored or caused the album's sale to be restricted. I'm looking for covers other than those ones, which means at least 99 percent of them.) I think you'll find album covers are described either as images, or not at all, because describing them as text would sound ridiculous; the same applies in Wikipedia as it does anywhere else. -- Mpt 19:41 15 Jun 2003 (UTC)

I don't think you want to appeal to common practice amongst online encyclopedias, Mpt - I've just looked at a random article on encarta.com which used "Leech" as alt text for an image of a leech.

I just looked at No Dice in the Lynx web browser. It's not ridiculous - it's a bit clumsy because of the description appearing at the top of the article. What is a problem is that the description appears as a link, but there's no indication of where the link goes. Martin

Balance in articles

Mpt says "the article would need to be considerably longer before the insignia would be worth mentioning" in response to one suggestion, and makes similar points elsewhere.

Articles in Wikipedia are always works in progress. It's entirely reasonable for a short article to mention a specific detail that might seem out-of-place in an article of that size. This is because every article is a work in progress, and what seems out of place in June might be completely reasonable in December.

Some people do think we should balance parts of a page, but that's only really supported in cases of massive imbalance - not for single sentences such as we have here. We tend to shy away from removing factual, verifiable information on grounds of "irrelevance". There are a few fringe cases where we do, but they're few and far between.

I fully agree. The "it's not in the article" argument (and anything like it) holds no water in an open-content environment. Nobody is claiming that every article on Wikipedia is devoid of out-of-place sentences, minor imbalances, and bias towards a particular thing (such as sheep lice, or a particular Grace Kelly movie). If my copy-editing is so lame, Mpt, why are't you helping us out by fixing it? -- Wapcaplet 00:37 16 Jun 2003 (UTC)

It's not that your copy-editing is lame (far from it, it's excellent). It's just that you're making the articles sound (or, in Lynx, look) like they were badly copy-edited, because the repetition in the text seems like an unintentional mistake.

If I reverted all your recent alternate text changes, I'd not only be pissing you off, I wouldn't even be helping you understand alternate text, let alone helping all the other Wikipedians. I'd rather figure out how to get everyone, or at least most people, to understand this (regrettably difficult) skill, and you seem like an excellent test subject -- because you disagree with me, but you have an open mind. -- Mpt 13:35 16 Jun 2003 (UTC)

I'm afraid I don't understand; mentioning the cover of The Swan in the article on Grace Kelly, the coin in the article on Diocletian, sheep lice in Lice, etc. cannot be justified with "Because I happened to have an image of it," but the stamp depicting Nero Wolfe, or the Stop Esso campaign poster are important and unique enough to get a sentence of alt text? Why aren't the bits you suggested for U.S. presidential election, 1789 and U.S. presidential election, 1900 important enough to be mentioned outside the image alt text? What is your criteria for determining whether an image merits having non-empty alt text? If it's mentioned in the article, then you say it shouldn't get alt text because it would be redundant; if it's not mentioned in the article, then you say it shouldn't get alt text because it apparently isn't important enough to be singled out for attention.

Wikipedia is open-content. Maybe you should consider starting a text-only Wiki.

-- Wapcaplet 20:31 15 Jun 2003 (UTC)

Very good questions, somewhat spoiled by your final line ("open-content" and accessible are obviously orthogonal, not mutually exclusive).

You misread Wapcaplet. Wikipeda is open content. Hence, you have the "right to fork". Hence you might consider starting a text-only Wiki, using Wikipedia content as a base. "open content" relates to "starting", not to "text-only". Martin 13:47 16 Jun 2003 (UTC)

I mostly threw it in there because it's the usual canned response to anyone with a beef about how we do things here. But yes, that's exactly the sense that I meant it in. -- Wapcaplet 01:40 17 Jun 2003 (UTC)

I once had a one-volume Macmillan Encyclopedia in which (IIRC) the Elvis Presley article was only a few hundred words, but included the fact that Presley featured on a US postage stamp. I judged it was similarly interesting for Nero Wolfe to have featured on a San Marino postage stamp, especially since he was a fictional character (as opposed to the Diocletian example, where putting emperors on coins was S.O.P. and not worth mentioning for any emperor in particular). If, however, Wolfe has featured on postage stamps for other countries too, then that should be mentioned in the non-alternate text (because the San Marino image by itself can't possibly express that); and of those stamps the San Marino one would no longer be particularly remarkable, so its alternate text would become "".

The sentence "Stop Esso France was sued by Esso, who alleged falling profits because the campaign used a parody Esso logo featuring dollar signs in place of the letters 'ss'." seemed incomplete to me -- I was thinking "used it for what? what was so profit-reducing about it?". It happened that the brochure worked as a good textual example of how the Campaign were disseminating the logo. If, however, anyone can find a better example but don't have an image for it, that example should go in non-alternate text, with the brochure's alternate text becoming "".

It seems very straightforward to me that a summary of how the states voted should be included in an article about a U.S. election. It just so happens that the images present such a summary even more clearly and quickly than text could, so the text should be alternate to the images -- only used if the images cannot be seen.

An alternat(iv)e approach

Obviously I need to find better ways to explain alternate text, but from the Schröder example and the election examples I think maybe I see the disconnect we're having here. The key is not whether something is "important enough to be mentioned outside the image alt text". It's whether it's important enough to be in the text, of an article of that length (modulo questions of balance), at all. Remember, in Lynx and other text-only browsers, alternate and non-alternate text are largely (and correctly) indistinguishable from each other.

Perhaps I've been trying to explain it completely backwards. Here's another way of looking at it, based on the "Augmentative authoring" approach of Jukka Korpela.

Write/edit the article text so that it is as complete, sensible, balanced, and well-structured as you can make it, as an image-less article, for your text-only readers.
If you have an image which is a clearer and/or quicker way of expressing something which you'd expressed with some of your text x, use the image to envelope that text -- no more and no less -- as its alternate, without the text itself changing its position within the article [[Image:...|x]]. Do that with as many of your images as you can.
If you have images which are relevant to the article but do not make any of the existing article text redundant (as is usually the case), insert them with alternate text of "".

That way you end up with a great graphical version, and your text-only version is identical to the best possible text-only version you produced in step 1. Win win. (Another advantage of this approach is that it works just as well for hyperlinks, audio, video, etc as it does for images.)

Do you think that would work as an explanation? Did it work for you?

-- Mpt 13:35 16 Jun 2003 (UTC)

Problems:

We typically add images to articles while they are still incomplete, nonsensical, inbalanced, and/or badly structured. We never reach the stage, with any article, of thinking "right! That's the text done!"
As you know, we currently can't use mid-paragraph images.

That said, I could support the addition of some advice such as: "One way to create alt text for new images is to move (not copy) text from the main text to become part of the alternate text for your new image". Martin

I don't know if the above is much better than the other arguments you have presented. It has the strong drawback of presuming that the text will ever reach anything like a "finished" state, as Martin notes; it also bears more resemblance to a guideline for when it's appropriate to submit images, rather than a guideline about how to write alt text for existing images. A guideline which says "Don't submit images for articles unless suitable alt text already exists in the article, or unless you are prepared to improve the article's text to create such suitable alt text" seems very likely to scare people away from submitting images at all, and would be quite counterproductive.

There seem to be many different perfectly good reasons to inser an image into an article. Possibly the most likely is "I happened to have an image which relates to this article." But there are also reasons like:

A picture is the only way to explain something (the photo of George Washington, or the Mona Lisa)
A picture is the best way to explain something (the graph of the Natural logarithm, or the map of Switzerland)
A picture is the easiest way to explain something (a picture of an Amiga computer or the Golden Gate Bridge)
A picture is the best substitute we have for poorly implemented or not fully standardized technology (such as any math article using TeX markup, or any images that were converted from SVG)

I am still not sure what I think of the best way to handle each of these. I am not fully convinced that the alt text must necessarily be a phrase taken directly from some (possibly hypothetical) all-text version; In some cases, especially in situations where no alt text will do (if the image is the only way to explain something), it seems acceptable to use some form of delimited descriptive phrase, such as alt="[Mona Lisa]". As you have said, the user may be interested in downloading or printing the image, and providing them with something indicating what it is they're downloading would be helpful (and, as noted on the other talk page, having null or empty-space strings used as hypertext links is probably not too wise). -- Wapcaplet 01:40 17 Jun 2003 (UTC)