(UPDATED) Acclaimed documentarian Errol Morris trades on – maybe trades in – his acclaim and purports to conduct an experiment measuring which typefaces confer credibility. Morris must be on the right track because he’s famous, must he not?


Morris doesn’t even know what he’s testing, or how

Ostensibly the comparison test involved a single paragraph of text in an article in the natural home of an A-list auteur like Morris, the Times. That graf was presented in one of six typefaces while surrounding text stayed constant. Then the same poll was offered to all readers.

Is Morris testing typeface, or typeface, measure, leading, colour, glare, viewing distance, and visual acuity? Actually, all the latter, plus screen resolution, screen orientation, operating system, and display mechanism (presence and type of antialiasing). Plus ranging versus lining figures. Plus ligature substitution.

I’ve read a dozen peer-reviewed papers on typeface testing over the years and easily a dozen and a half unreviewed papers, and I can confidently tell you that nearly all of them make the same mistake Morris did – not just failing to control for real factors involved in reading text but acting completely unaware of them.

We aren’t dealing with just one kind of screen

It is preposterous to suggest we’re testing the “same” typefaces when we don’t know what screens the subjects are using. We can be confident some readers viewed the experiment on a CRT under Windows with no antialiasing, and, at the the other end of the evolutionary scale, on iPad 3s and MacBook Pros with retina displays. Do your own test: Walk your iPad over to a Windows shitbox in a second-rate office and look at the same text, from any source, side by side on both screens. Is it really the “same” text?

It’s an insult to the intelligence even to have to state these facts. You have to control the reading environment or you aren’t testing anything, let alone reading.

(Morris’s programmer/researcher, Ben Berman, didn’t answer the E-mail I sent.)

Did the test even work?

Unlike apparently everyone else who covered this non-event, including otherwise smart people, I studied the source code. Markup (EM with a class) was used with some JavaScript and real Webfonts. (I doubt Morris and the Times had have a valid licence for all the fonts used.) So it should work, should it not?

Well, did it? Was the investigator looking at the same screen as the subject to ensure the subject really was reading a specific font? No. Why are we even reporting the results of an experiment that experimenters were never present for? (Because it’s Errol Morris and he’s famous for something else?)

Computer Modern is a dead giveaway

There’s a special breed of twit who thinks Computer Modern is a real typeface that is really used by real people. Its inclusion in this so-called experiment speaks volumes about how little Morris knows about typography, and how much starfucking he’s willing to do. (Then he expects the same treatment.)

I remember the origin of this typeface because, as a teen in New Brunswick, I phoned up Donald Knuth after I read his articles about it! I never got him on the blower, because even then he was much too important, but I distinctly recall his secretary’s correcting my pronunciation of Knuth. (As with Knopf, pronounce the K.) Computer Modern, a cœlecanth from an antediluvian computing environment, is as much of a relic for its discipline – scientific typesetting – as Adobe’s spindly digitizations of centuries-old metal type are in general. And both of them are found in this same “experiment.”

Worse, the inclusion of Computer Modern recapitulates a typical error of typeface “testing” – using an implausible ringer typeface, one that nobody with type knowledge would actually choose, as a claimed control. In extreme cases this sort of thing approaches scientific fraud.

Surprise: A centuries-old Establishment typeface “won”

People who know a little about typography know what Baskerville is. A little knowledge is dangerous. What they don’t know is that Baskerville works solely as a hot-metal letterpress typeface and poorly in every other means of reproduction, including onscreen type.

It borders on impossible to get people to believe this, if only because most weren’t around in the 1980s when Adobe did a half-assed job digitizing old letterpress typefaces. Those of us who lived through that era had a direct before-and-after experience, with access to beloved hot-metal typesetting and the worst knockoffs typography had ever known. As with the iPad/CRT test I mentioned above, we could put these works side by side and compare them. And as with Courier, we still suffer from those early mistakes. (UPDATE: Albert-Jan Pool notes that PostScript was merely the newest step in a sequence of paltry digitizations, one building on the other.)

Digitized letterpress typefaces are too skinny and faint when presented in offset printing or on screens. The effect is more pronounced with transitional typefaces like, yes, Baskerville, where vertical stress seems heavier than it really is and vertical strokes seem even spindlier. But the original letterpress Baskerville, still easily found in old Penguins and other books typeset in hot metal, presents a pleasant softness – exactly the effect Zuzana Ličko aimed for with Mrs Eaves. (Errol Morris knows less than half as much as he needs to on every topic in this discussion, so it doesn’t surprise me that his 4,800-word disquisition never mentions Mrs Eaves.)

If you somehow doubt my claim that Baskerville works only in its native medium (hot metal), read any recent book typeset in that face. My library copy of Mark Leyner’s The Sugar-Frosted Nutsack (2012) is indifferently set in this Baskerville manqué, with too-close letterspacing and too-wide wordspacing (“spaceband”). (Plus the type size is too big; Baskerville sets narrow but tall.) The bold is more like a black and is a post-facto joke.

There’s no designer credit, so I couldn’t verify the following, but I am quite sure this is merely an OpenType transcoding of a PostScript digitization. (That at least is a step up from typical book design, which involves the use of the same Adobe Type 1 font files – with “expert sets” – that the designer bought in the 1990s and never upgraded.)

If you can’t be bothered to look at a real book, inspect the high-resolution scan I uploaded. (I have a 48 MB TIFF if you want it.) Still think Baskerville really works in offset printing or on screen?

So how was it the “winner”?

A Snopes-quality failure

This experiment, though worthless, is so deliciously quotable by a certain chattering class that it will surely take its place in the mythology of reading alongside the debunked “study” that we can breeze through a sentence even if every word had its letters scrambled. The same filmmaker who snowed his viewers with irrelevant detail masquerading as research – check the animated sequence on EXIF data (!) in Standard Operating Procedure – will join the rogues’ gallery of propagators of fun facts about typography that may be fun but never were facts.

Errol Morris will keep right on doing what he’s doing. Why, just a few days after publishing this abomination, Pentagram proudly announced it had designed his latest hardcover book (and movie trailer and Web site), ironically entitled A Wilderness of Error. Morris will keep right on failing upward.

Actually, come to think of it, the Times is just the place for rich, powerful, well-connected and approved people to publish falsehoods.

The foregoing posting appeared on Joe Clark’s personal Weblog on 2012.08.26 14:13. This presentation was designed for printing and omits components that make sense only onscreen. (If you are seeing this on a screen, then the page stylesheet was not loaded or not loaded properly.) The permanent link is:

(Values you enter are stored and may be published)



This personal Weblog is unlikely to be updated again until my next book comes out. (See Best postings)

Other reading

Topics of interest

Typography ⁓ graphic designTTCCanadian EnglishInversion

Archives by category

Archives by date

Just add /year/month/day/ to the end of site’s URL, You can add just /year/month/, or just /year/, if you wish. Years are four-digit, month and day two-digit (with padding zero below 10). For example:

Copyright © 2004–2021

You enjoy

Transgenderism is to be opposed categorically