The second poster above took a brief look at one of the pictures and decided this is an exercise in graphic arts. Not so, the icons are all automatically generated. (Please take the time to read the abstract).

There have been usability studies that prove the exact opposite of what the article states, that icons/images are quicker to understand than language.

Actually the article says that icons/image are quicker to search and are more easily remembered than text labels. It does not refer to "understanding".

The article also does not propose putting labelstooclosetoeachother. Rather, it shows that,

if all other things are equal (spacing, icon size, text label, etc.), people perform better if the icon is distinctive than if generic,

Some of the pictures are high resolution; these assume that in the future there will be more resolution, or zoomable icons (OS/X has this somewhat already). But the user studies were done with standard 64x64 icons.

Sorry if replying to the main thread is in bad form.. there is unfortunately so little forum activity on nooface.

I read the entire article and I understand what he meant. I did not take a brief look at anything, in fact it is kind of ironic that you say such a thing (that I looked only at the image) about someone who is commenting on how text based approaches are superior.

• This is an exercise in graphic arts even though it is automatically being generated. That doesn't make much sense honestly, most effects in computer aided graphics are automatically generated (at the request of the user/designer).
• Understanding is integral to usability. The usability studies stating that icons are quicker and easier to remember are ignoring the fact it takes extra time to recall the meaning of the icon versus our near instant recall of natural language. Images take longer to recall than language.. this is a fact. Recalling specifics of a small icon (64x64) like one line is more square or dotted than another is even more difficult.
• He mentions using generic icons versus specific icons and text labels for icons. He does not study the difference between having just text and no icons. Of course there is extra space taken up by having a text label in addition to an icon. This space not only effectively reduces the available display size of content, but also makes it even harder to directly focus on the text itself if you just happen to forget or not even know what an icon means. This last case will happen even more with automatically generated and assigned icons.
• Zoomable interfaces are here now and are just as applicable to text as they are icons or other graphics. In fact, currently the most in use software that features zooming can only zoom text (Mozilla*/IE). I know Opera of course can zoom everything and that is dandy. I don't know/think it is the same however since when you 'zoom' text it should stay within the confines of the window or element.

(*&_\) - Docs
  Docs

Which of the above is easier and quicker to recall or use? Here, compare the following two images: Iconic - Textual. Please no mention of the mouse cursor being an icon. Yes it is, and an icon is necessary for pointers to work. I of course use a keyboard to select links since that too is quicker. Imagine now if every link had a number automatically generated for it. When you hold down a meta key and start typing numbers then release it a link will be selected with that number and the system will do whatever action is specified for that link. The number links could show up when you hold down the meta key or could be toggled with another meta key so they stay on fulltime (and hence your text view stays consistently formatted). That trully would be fast! You can almost do something of that sort with elinks.

You can check out the book The Humane Environment (THE) by Jef Raskin and his website at humane.sf.net. He has a good deal of usability study references. I don't agree with everything he says and his book has considerably more info than his website.. though he has lots of good ideas.

You might be more persuasive by giving some evidence, especially since there is a lot of evidence pointing in the other direction:

A Paivio, Pictures and words in visual search. Memory & Cognition 2, 3, 515-521, 1974.

R. Shepard, Recognition memory for words, sentences, and pictures. J. Verbal Learning Verbal Behavior 6, 156-163, 1967,

More recently,

Jerymyn et. al., The Design and Analysis of Graphical Passwords - 1999

"Cognitive scientists and psychologists have shown that there is a substantial improvement of performance in recall and recognition with pictorial representations of to be remembered material than for verbal representations.
...
Studies exhibiting strikingly high differences in memory recall of pictures over words have since been replicated on numerous occasions [She67, Sta73, NRW76, BSH77]."

And lastly see the two studies reported in the paper under discussion, which cover both search speed and memory for the invented icons.

The advantages of appearance over text are also evident in daily life --
books and cars are two types of object that have both a textual label (title or license plate) and an appearance. When you look for a particular book on a bookshelf, you don't scan titles one-by-one. You _could_ do this, but it's much faster to search by appearance, and then verify the title if there is any ambiguity. Likewise, you don't look for your car in a parking lot by examining each and every license plate.

Similarly, memory for appearance seems intuitively to be better than memory for words. I have a lot of books, and while I can always remember the appearance of a book, sometimes remembering the title takes a moment of concentration, or worse, it is forgotten. The converse, remembering the title while having no idea of the appearance, seems unimaginable. It's not just me; in my experience, people commonly say things like "have you seen that little red travel guidebook for England", indicating that they know the book's subject and its appearance, but they've forgotten the exact title.

He does not study the difference between having just text and no icons.

This has been studied by other people, e.g. the Byrne reference in the paper. In summary, it is questionable whether generic icons help at all. However, if there are several classes of generic icons, e.g. distinct icons for a Word file versus an Excel file, Byrne states that people make use of this, by first scanning for files of the desired type by appearance, then searching the filenames (similar to the way we look for books or cars).

(I also apologize for posting this response at the top level, but the whole discussion got started at this level so I'll continue here).