Replies: 1 comment

Actually their research -- as well as their other studies -- tracks quite well with the typographic rules-of-thumb from my graphic design days.

Verdana (as well as Georgia) was specifically designed by Mathew Carter for on-screen usage, so it's not surprising they were among those perceived as most legible. The size of the lower-case letters is huge -- a large "x height" -- and the interior white space of the letters -- the "counter" are relatively large and open.

While Comic Sans and Courier share these characteristic, although they probably weren't designed with them in mind. Comic Sans is also relatively bold, providing the strongest contrast, which makes it clearly legible -- although the contrast would become tiring at longer lengths (the passages were only 1024 words.)

These quite noticeable in the side-by-side samples A good counter-example is Agency, which condensed to the point where it's tough to read even in print -- it would only be used as "display" type (headline, pullquotes, etc.) and never as "body" type (the main text) -- because the "counters" have closed-up and the letterspacing is so tight that it can be difficult to resolve where one letter ends and another starts.

Not suprising Verdana was people's preferred typeface, followed by Arial and Comic Sans, which both have a relatively large x-height compared to the others studied. They're also sans serif fonts, which tend to be more legible on-screen because the "feet" of serifed types typically renders poorly in the low-resolution environment of on-screen, making the "feet" disproportionately large and makes the letterforms harder to discern.

Courier is more difficult to read for a different reason. It's "mono-spaced" -- each character is given the same amount of horizontal space regardless of its actual width -- it's harder to resolve individual letters into words, consequently text in Courier took almost the longest to read -- equal to a "calligraphic" typeface that's ill-suited to on-screen usage.

It's not surprising that Times, which was originally design to cram as many characters into a newspaper column and was one of the worst typefaces that could've been picked for on-screen usage, was rated the least readable and took among the longest to read.

Given the small sample size (22 people) I'd take the "personality" ratings with some caution. Both the top-rated Bradley and Corsivia were "callagraphic" typefaces, but Bradley (intended to look like handwriting) is hardly something most graphic designers would consider "elegant."

Courier was underdoubtedly perceived as "business-like" due to its origin as _the_ typeface used on manual typewriters (an interesting example of how specific "looks" outlive their orgins -- since most of the participants were probably too young to actually remember manual typewriters....). Times was likely perceived as business-like because of its ubitious usage as the "new Courier" for laserprinters.

Comic Sans lead the pack in being "youthful and friendly" because it's somewhat callagaphic without being being overtly so (it's actually an imiation of "Tekton," which was based on an architect's handwriting). Comic Sans also has intentional "off-balance" elements (look at the different curves in the "S" and the angled cross-bar of the "e") that are perceived as adding energy. (I won't bore you with the perceptual pyschology theories behind this -- nor are they theories that most designers are consciously aware of, they just know these things from training and experience.)

While it's good to have this tested in on-screen environment, a lot of these principles are well-known (and well-tested) in the graphic design field. Unfortunately, a lot of CHI/usability people don't seem to bother to talk with graphic designers about it. (Note: I've got no idea whether the researchers did or didn't....but none of their studies make any mention of previous work along these lines.)

Finally, a useful critique from the SIG-IA list where this is also being discussed: the study didn't look at the context of the task that involves reading (they just looked at linear reading of natural language in an extended text). So for example, Courier is better for reading programming code, precisely because it tends to focus attention on individual letters rather than whole words.

To quote Andrew McNaughton (who raised the point):

>For a high proprtion of web usage, the important metric is not how fast a
>person can read a passage, but how quickly and accurately they can pick
>out key terms that relate to what they are looking for without having to
>read everything. It's important to remember that most of the
>text on most pages doesn't get read by most people, and it's certainly not
>read in a linear fashion. Kind of a 'randomn access' process really.
>Capitals that stand out and space around the words is good for randomn
>access, but not I think for linear reading.

This is fairly different from the decades of legability studies that have been done for print.