Errol Morris' quiz about killer asteroids was a secret experiment to find out how fonts affect our thinking

Remember Errol Morris' New York Times quiz about being an optimist or a pessimist about killer asteroids? It turns out the quiz wasn't really about being an optimist or a pessimist. It was about how fonts affect people's thinking. The quiz was presented in five different fonts, and the frequency of the answers were compared to the fonts in which the questions were asked.

We all know that we are influenced in many, many ways -- many of which we remain blissfully unaware of. Could fonts be one of them? Could the mere selection of a font influence us to believe one thing rather than another? Could fonts work some unseen magic? Or malefaction?

Don’t get me wrong. The underlying truth of the sentence “Gold has an atomic number of 79” is not dependent on the font in which it is written. The sentence is true regardless of whether it is displayed in Helvetica, Georgia or even the much-maligned Comic Sans. But are we more inclined to believe that gold has an atomic number 79 if we read it in Georgia, the font of The New York Times online, rather than in Helvetica?


Results from the asteroid quiz:



Morris concludes: "Comic Sans has the lowest rate of agreement, and one of the highest rates of disagreement."

Hear, All Ye People; Hearken, O Earth (Part One)



  1. These are the most blatantly dishonest bar graphs that I’ve seen in a long time.  15627 is 95.5% of 16359, but the Comic Sans bar is shown as about 1/3 the height of the Baskerville bar.  Drawing these graphs honestly would reveal that the differences are tiny and, depending on sample size, possibly not even statistically significant. I expect better from Boing Boing.

    1. It’s the default way Excel plots these values (min value is 15,000 for the top chart, for instance). If the person doing the graphs then does not think very hard abut how appropriate this representation is to the data actually being represented, you get this kind of silliness.

      EDIT: Worse still, the default Excel plot includes the y-axis which makes the non-zero minimum clear. They decided to get rid of the axis and just display the values above the bars, which resulted in an even greater misrepresentation. Sigh.

      1. Yes. This is particularly annoying because despite the efforts of the more computational savvy biologists to get their bench colleagues to use things like R, which actually plot things correctly, Excel is still the standard graphing tool used by most bench biologists.

    2. Baskerville was signficant.  Cutting axes, although should be noted, is a fine way to highlight said statistically significant differences.  Showing the bars identically would be just as midleading.  Normalizing the graphs would have been another way to highlight the differences, but that would be even more confusing for a lay reader, I would think. 

  2. “Computer Modern”! whoohoo…. hail hail [la]TeX, land of the brave and free! and blessings unto Saint Knuth.

  3. Someone threw together a nice “adjusted” version of the graphic, to make up for the misleading scale:

    Also, since I’m sure someone will ask why it is that people hate comic sans so much, here is an article and talk about that very subject:

  4. I hear that people who use monospaced fonts are perceived as highly intelligent and incredibly sexy.

    1.  This amuses me greatly, because when I first read this font, the comment glitched and cut of after “…perceived as…”


  5. Back in the late 1970″s, I attended a printing workshop at the literary press, Copper Canyon Press.  Co-publisher Tree Swenson explained that serif fonts are easier to read than san serif ones, because the “wings” help the eye move from letter to letter.  I never forgot her words when I decided to finally complete my college education in the 1990’s.  You see, I noticed that my younger classmates preferred san serif fonts, so I figured I could give myself a leg-up by doing my professors a favor and make my papers just a little easier to read by using Times New Roman.  I’d like to think it made at least a little difference.  

  6. I used to tell my college students to use serif fonts on their papers, although if they got snippy about it I’d laugh and say they didn’t have to — I just preferred serif fonts.  I think they thought I was crazy (for that, and other reasons too).  I was just being honest.  I hate reading anything lengthy in a sans-serif font.

      1. Yet when I handed in my assignments in Wingdings, they marked it as a fail. Everything in moderation, I suppose.

    1. The rule of thumb I had always heard (Adobe’s “Stop Stealing Sheep, and learn how type works”?) was serif for under 24 points, sans over, so use it for headlines, etc.

  7. If I follow correctly, for this study n=157,120. For Comic Sans 60.4% correctly agree with the statement. For Baskerville 62.8% correctly agree. That’s defiantly statistically significant with a caveat that the population is only Washington Post Opinion Page (or Boing Boing) readers. That 2.4% difference could be important to you if you are running from president or something?

    1.  If I’m running from president, I’m probably thinking about how to avoid the secret service agents rather than anything about statistics.

      1. But you want your the signs you leave behind saying “He went thattaway” to be believed, right?

  8. I often think of “verbal fonts”, accents, and how easy it  is for someone with a posh British accent to sell us Americans trash. Now think of other non sexy accents we would have no problem hanging up on. 

  9. What do people have against Georgia?  It doesn’t look that different from Baskerville, which is at the other end of the scale.

      1. It’ll have its day again. We’ll all get on that midnight train. Sometimes we all crave a simpler place and time.

      1.  All serif fonts look approximately the same to me.  When the UofC press published something of mine some years ago, I remember talking to the designer, thinking he was seeing something I couldn’t, wondering if I had the typographical analogue of color-blindness.

        Given the care with which typographers do their work, I don’t doubt that there is a great difference.  I know that certain editions of books please me more than others, engage my eye better, allow me to read longer.  Maybe it’s the font or the reflective property of the paper or how much the weight of the book fatigues my hands.  I don’t know.

  10.  I teach ESL and some of the resources I pull off the internet have worksheets in Comic Sans.  My students and other teachers have complained that the font is too hard to read.

  11. Loftus, E.F. & Palmer, J.C. (1974). Reconstruction of auto-mobile destruction: An example of the interaction between language and memory. Journal of Verbal Learning and Verbal Behaviour, 13, 585 -589

  12. Fonts remind me of how much I hated typography class and regretted not telling the teacher to f_k off when he told us to do fifty thumbnail sketches and then bitched about people doing exactly fifty thumbnail sketches.

  13. I find that I have an excess of trustworthiness and credibility. 

    I’m switching to Comic Sans posthaste!

  14. Someone in the comments on the Times site pointed out the important caveat that Baskerville is not a web-safe font and, as a result, many who completed this quiz didn’t actually see Baskerville at all.  Can a web designer reply and indicate what font Baskerville would be rendered as on a Windows machine?  Is it Times New Roman?  If so, this effect is probably driven by familiarity with Times New Roman.

    Would be great to see the results adjusted to only include those from configurations that would likely render the text in Baskerville itself.

    1. And there are people like me who have deleted Comic Sans from their machines and who don’t have Computer Modern installed. Was that kind of thing factored in to the study?

      1. It’s possible they did the questions as images (or webfonts, but I don’t think you can rely on 100% support rate for those yet). 

    2. It’s possible – can’t be sure without seeing the original quiz – that they embedded the fonts using @font-face, so the viewers would have downloaded them from a web server rather than relying on them being installed. That’s certainly how I would have done it.

    3. Typically it’s specified in the css fallback rules. You might say Times New Roman is the first fallback if Baskerville isn’t found, and then a generic serif font which the browser will pick for you based on it’s default font setting. The css for that is something like “font-family: Baskerville, Times New Roman, Serif;”.

      If you don’t do it that way, then the browser will figure it out for you based on it’s default serif setting anyway, which I believe is Times Roman on OS X and Times New Roman on Windows (it’s  Helvetica and Arial for sans-serif if you’re curious).

  15. The original graphs from The NYTimes are weighted by strong/somewhat/slight levels of agreement levels for each font in the graphs shown here — I think that explains the wide gaps visually here, doesn’t it?  The unweighted graphs are presented first in the original article.

  16. so people are inclined to believe a statement presented in a similar font to a textbook, but are inclined to question a statement presented in the same font as passive aggressive work notes about making sure you mark your food in the fridge? truly shocking

  17. This drives me mad.  This is about typefaces, not fonts! These words are not interchangeable.  ‘Typeface’ describes the graphical style of the letters.  ‘Font’ describes a combination of typeface and size.  I would really, really expect someone conducting research on this subject to know this.

    1. #FirstWorldProblems. 

      Not to brush off the fact you are correct, this is a very minor thing. Chill.

      1. Would you accept font-type, perhaps? That seems to be significantly more common nomenclature than typeface.

        1. I’m not that fussed either way; it just seems odd, to me at least, that people get mad over such things.

    2. Microsoft Word lists these ‘typefaces’ under the “font” tab.  In fact, on my screen right now, I see “recently used fonts”  followed by “Times New Roman”, “Ariel”, etc. This is how the term is used now, sorry.

  18. Random factoid: the “Canada” wordmark is in the Baskerville font with a modified… uh, I think it was a modified ‘r’ to wit.

  19. What Morris’s experiment doesn’t test is people’s response to how each typeface is used: i.e., typography. That’s at least as important as the choice of font, in how we perceive and respond to any text.

  20. Big news. As any decent graphics designer will tell you, there’s indeed an emotional quality in any typeface. Doesn’t matter where it comes from – mostly it’s just “this is always used by banks, so it’s a perfect typeface for another bank” – but this is why chosing the right font for a corporate identity can be a very hard thing to do, up to the point where they need to design a completely new typeface.

  21. I don’t see how being the 3rd from the bottom in terms of rates of disagreement makes Comic Sans have “one of the highest rates of disagreement.” It looks like one of the lower rates of disagreement, to me.

  22. I <3 Baskerville, but my true love is Garamond.

    On a completely tangential note, an especially convincing font was part of Doofenshmirtz's episodic scheme to take over the Tri-State Area in the recent Phineas & Ferb episode "Ferb Latin."

  23. Only Baskerville came out as significantly different from the others, not comic sans.

    As for the y-axis Nazis, there is nothing wrong with cutting the axis to show the difference, which was significant.  Highlighting their sameness with a full axis is even more misleading.  Yes, he should have drawn our attention to the axis switch, but he also could have normalized the data to a single font (say, the font the intro to the quiz was in) and highlighted the differences that way.

    Blindly following Tufte off a cliff….

  24. And of all the Tufte zombies complaining about the bar graph, not one pointed out the ugly, needless pie chart in the middle of the NYT article.

  25. Would it be considered ‘mean’ to go IRB shopping to see if I can find somebody who will go full-metal-hardass over this unapproved human experimentation?

  26. I have to say, I’m shocked there aren’t greater differences between fonts than this.
    As a designer, I was taught that your typeface is for design as your clothes are to you. A comic sans would be a sloppy T-shirt and shorts, while a helvetica would be a modern cut suit.

    That said, I guess in north america most people don’t give a damn about how you dress either.

  27. People have come to use the wrong typographical term for “typeface” or “face”.

    “Font” refers to a particular style within the typeface such as outline, bold italic, light condensed, etc.

    The terms “font” and “typeface” are not technically synonymous.

    1. Isn’t that just a function of the fact the Microsoft Word calls these “recently used fonts” when you click on the “font” tab? 

      Almost everyone on the planet with a computer sees these called “fonts” every day. 

  28. This would be much more interesting if they were able to correlate some feature(s) (ex. Surface area) of the typefaces with agreement and disagreement.

    If there was a correlation with surface area, one might make the prediction that agreement could be altered by changing the font size.

  29. So is this post a meta fakeout?  That is, it this story — which reveals that the real issue is NOT about asteroids but about fonts — REALLY another layer of deception and the REAL issue under review is the effect of horrible bar graphs on readers?

    1. What’s wrong with the bar graph?  It had to be cut to show the statistical differences in the data, a legit practice.

Comments are closed.