Baskerville bulls—t?

Chris Dean's picture

After having the same article tweeted, facebooked, emailed, posted, mentioned to me, and a few days trying to piece together a messy collection of hyperlinks and pages, I prepared the following article review. And as far as bullshit goes, a good scientist can’t say that. It was just a hook. In order to support or refute the claims of this study you would need access to the same materials and data. Have at it Nick. I know how much you love the role of science in the practice of typography ;)

In one sentence
A short passage set in the typeface Baskerville was found to be more believable than the same passage set in five other typefaces.

APA citation
E. Morris. (2012, August 08) Hear, all ye people; Hearken, o Earth (part one) [Web log post]. Retrieved from

What did they change?

What typefaces did they use?
Computer Modern
Comic Sans

What did they measure?
Reader’s agreement or disagreement with a statement about a passage.
Reader’s confidence with their agreement or disagreement.

Who did they test?
Readers of The New York Times blog: The Opinion Pages
~10,000 people clicked on the article.
~50,000 people took the quiz.
45,524 data sets were used for statistical analysis.
No demographic data was reported.

Experimental design
Unknown. It is not reported if subjects took the test more than once — “If you have taken it before, please take it again.” (Morris, 2012). As such, a repeated-measures mixed-subjects design is most likely.

Why did they do this?
To determine if typeface influences the believability of a passage of text.

What did they do?
Subjects read a 368 word blog post on The New Your Times blog, The Opinion Pages. Within the post was an 80 word passage about an asteroid’s near collision with Earth. Through use of a computer program, the passage within the post was presented in one of six different typefaces: Baskerville, Computer Modern, Comic Sans, Trebuchet, Helvetica, and Georgia. At the end of the post readers were asked two questions; One, if they agreed or disagreed with a claim in the passage “…we live in an era of unprecedented safety…” and a second, asking how strongly they agreed or disagreed with the claim.

What did they find?
Subjects found the passage set in the typeface Baskerville to be more believable than compared to the other five other typefaces.

What does this mean?
If the simple act of taking the time to choose an appropriate typeface has the potential to influence readers’ perception of your credibility, perhaps it would behoove you to case the joint before you B & E.

Despite the number of datasets analyzed, given that the test was administered online, it is not possible to determine how many individual actually actually took the test. Especially given Morris’ (2012) request in the original post “If you have taken it before, please take it again.”

The passage itself is also very short at 80 words. This and there is the possible confound of an interaction between typeface and content. For example, if Comic Sans could be the winner if the passage was about clowns.

In addition to this, objective measures of human performance such as time to read, and comprehension accuracy, were not reported, making it difficult to determine the validity of subjects’ responses.

Of special note, data analyst David Dunning of the Department of Psychology of Cornell University admirably acknowledged several potential criticisms of this study in his statement “You are collecting these data in an uncontrolled environment (who knows, for example just how each person’s computer is rendering each font, how large the font is, is it on an iPad or iPhone, laptop or desktop), are their kids breaking furniture in the background, etc. So to see any difference is impressive.” By this he means that in an uncontrolled environment, the fact that everyone is using a different device, has a different education, speaks a different language &c could potentially nullify the effects of changing the typeface. In this case, it did not. And Dunning’s (2012) lay-person description of statistical analysis is one of the best plain-language examples I have ever seen. I look forward to reading follow-up studies.

You wouldn’t happen to have an extra $30,000.00 of grant money kicking around would you?


Further reading
Brumberger, E. (2004). The rhetoric of typography: Effects on reading time, reading comprehension and perception of ethos. Technical Communication 51(1), 13–24.


You can find the same review on my website at

Nick Shinn's picture

I’ve already had at this, here at Typophile!
Now if only the site were searchable, I could remember what I said…

Chris Dean's picture


Chris Dean's picture

Have you read Brumberger (2004)?

Si_Daniels's picture

>Have you read Brumberger (2004)?

Read it? I downed the whole bottle!

rs_donsata's picture

The test is biased to boost the sales of Baskerville since it was an unfair competition: none of these typefaces but Baskerville is ever featured in serious books. It's like having Clint Eastwood give a bad news, nobody can beat him.

Baskerville - Mr. Eastwood
Computer Modern - Nerd bookworm (knows too much)
Comic Sans - Will Ferrel (looney)
Trebuchet - Computer wizard (hacks emails)
Helvetica - Supermodel (knows nothing)
Georgia - Bossy teacher (manipulative)

dberlow's picture

A news article set in Baskerville was more believable to real world users than one set in Computer Modern, Comic Sans, Trebuchet, Helvetica or Georgia.

I'm thrilled, thanks for the great review.

Chris Dean's picture

It’s the methods that are meant to be evaluated, not the results.

Nick Shinn's picture

You mean there is a possibility the Pope might not be Catholic?

dezcom's picture

The Hounds of the Baskerville had a more frightening growl after eating raw tiger meat than they did when eating Purina Puppy Chow.

dberlow's picture

"~50,000 people took the quiz."

This is good.

"For example, if Comic Sans could be the winner if the passage was about clowns."

This would be good.

"It’s the methods that are meant to be evaluated, not the results."


"In addition to this, objective measures of human performance [...] were not reported, making it difficult to determine the validity of subjects’ responses."

I'd much rather have doubts about the subjects, than anyone else involved in such a study

Wouldn't you?

Chris Dean's picture

I’d much rather have doubts about the subjects, than anyone else involved in such a study.”

By this do you mean the fact that the population was not controlled?

While I agree, it’s an uncontrolled population, this does give the study a higher degree of external or ecological validity. And as far as typography + psychology studies go, 50,000 Ss is huge in my books. Good or bad, it’s certainly noteworthy.

Personally, I find the length of the passage (80 words) and the number, and nature of the questions, — 2 questions, 1 yes/no, 1 three point Likert scale — to be of greatest concern. Combine that with the design noise around the test portion, and it gets even messier. Below you see the full post (shown at 50% of what I saw it as on my machine) with a red line around test portion (the part where they change the font. The rest remained unchanged). IMO, that’s a pretty damn small portion.

dberlow's picture

"? you mean the fact that the population was not controlled?"

Not exactly. I mean I'd rather know nothing about the subjects than not to know the researchers have screwed up something fundamental typographically.

I would share your concerns for the surroundings and the portion, but that there is nothing about this test that claims it's about visually insulated long-passage immersive reading, is there?

But I think the most important thing about the study is as follows... About 25 years ago, Adobe tried to strip the postscript-using world of its free font choice. They had a couple 100 fonts and, via licensing, threats and technology, were claiming control of the type one format. This failed of course, everybody got to have any fonts they wanted and publishing bloomed, wherever postscript went.

Then about 20 years ago, the web came along and stripped the web using world of its free choice in live fonts, forcing anyone who wanted a typographic personality to do it via graphics, and all but eliminating typographic choice in reading sizes, unless bad versions of Helvetica, Times Roman and Courier, at low resolution, was the publisher's "thing."

Only 8 years ago did the healing begin on this with the move towards @fonface. Meanwhile a whole generation of developers (will get to "that generation" of readers in a second), grew up reading a small selection of fonts used for all purposes, and when this generation got to do their technology thing for reading devices, these devices dint have many fonts... Or not as many as the wide spectrum of fonts authors and publishers used in the printed books they were selling to readers...

So, after all that, for 50 k people pf that generation to say no thanks, I want my Baskerville?

Té Rowan's picture

I want my... I want my Basker-Vee...!

Look at 'em yo-yos! That's the way to do it:
Money for nothing and 'em fonts for free!

For some reason, I feel like slipping Dire Straits into the player.

Chris Dean's picture

Does anyone know the the thread that previously discussed this study?

dezcom's picture

I remember it but can't find it.

Here is one similar but not it.

Syndicate content Syndicate content