A topic that never goes away is fixed vs. fluid vs. “elastic” vs. “Jell-O®” CSS layouts. I just tire of the whole thing. Any kind of layout that does not use in, cm, m, pt, or pc units meets WCAG 1.0 Checkpoint 3.4 (“Use relative rather than absolute units”). That means you can use px if you want; unbeknownst to WAI, which can’t read its Consortium’s own specs, px is a relative unit.
Anyway, proponents of fixed layouts tell us that, for typical font sizes, a fixed layout at least lets you regulate the line length, which, as everyone knows, works best at around 72 characters. (Actually, even that number is debated and contingent.) Proponents of fluid layouts tell us you can use max-width in real browsers to make sure your line lengths stay sane, and point out that huge fonts produce too-short lines (also bad linespacing) in fixed layouts.
Well, hold on to your hats, kids, because long lines might not be so bad after all. An information-dense paper by Mary C. Dyson, “How physical text layout affects reading from screen,” summarizes much of the available research on line lengths, particularly in studies that looked only at onscreen reading but also in studies that (mis)applied print-reading research to the screen.
Dyson explains that most studies are poorly constructed, with many “confounding” factors that the researchers did not anticipate. You can’t just study line length, for example (which few papers did in the first place); you also have to consider modality (print vs. screen), font, linespacing, and resolution. (She forgets to mention visual acuity.) Nonetheless, the trend in the research she cites is reasonably clear:
Most of the studies on line length report faster reading with longer lines… Increasing characters per line, but maintaining a constant visual angle [another way of measuring line length], can result in faster reading… whereas differences in visual angle have only a small effect on reading speed within the middle range applicable to most displays.
This doesn’t mean people say they like longer line lengths:
In general, subjective judgements of variables relating to text format on screen have not been in close agreement with objective performance measures, such as reading rate and comprehension. Although longer line lengths may be read faster, people prefer a more moderate length. With columns, a single wide column is read faster, but narrow multiple columns are preferred.
(In onscreen reading, people will tell you they prefer things that are actually bad for them.)
Dyson’s research is not a justification for marathon line lengths. It merely indicates that the issue is more complex than in print typography, and, I infer, that categorical declarations are hard to back up with evidence.