The effect of typefaces on credibility

Those columns graphs were very misleading to me. Here is some adjusted image showing the whole image for easier grasping of the dimensions: http://i.imgur.com/QS8PA.jpg (I am not 100% my math is correct but a quick calculation in my head says the dimensions seem correct)

I'm not that great of a writer. My papers were usually in the B range and I almost never got an A. Then I started using LaTeX and now my papers are used as examples in classes.

An empirical anecdote: when I was in high school, all of my lab reports for chemistry and physics were typeset in LaTeX while most of my friends either handwrote the mathematics or used MS Word's Equation Editor. There were multiple occasions where a friend and I made the same mistake (we usually worked together; yes, we cited each other) and the deduction on my report was less than the deduction on his. It wasn't huge, usually -1 point vs -2, but there was consistently a difference.

When the survey initially came out I was randomly given Computer Modern. The day before, I had painstakingly converted Computer Modern it to a webfont for a friend's blog (http://www.krisjordan.com) and I was shocked to see it on nytimes.com.

I quickly went over to a different (Windows) machine to try it out because I couldn't believe my eyes. That one was given Georgia, so I mistakenly assumed that Errol Morris was such a type hipster that he included Computer Modern in his type stack if it was installed locally. It was pretty funny to see this today.

One thing I will say, is that the Computer Modern webfont they used is a disaster. It had tons of aliasing issues. I wonder how they sourced it since natively it isn't in a normal font format. (Knuth!) That certainly would skew the results.

How did they control for whether the fonts are actually installed on the participants' computers or not?

Also, did they control for desktops vs smartphones vs tablets? It's reasonable to hypothsize the device's screen (and zoom level on mobiles) affects typeface rendering and its perception.

All in all, intersesting and worthy of more work, but I want more to believe the result more.

I really hate those "Weighted Agreement/Disagreement" charts.

For Weighted Agreement, it looks like Comic Sans had a way lower agreement rate (it looks like 60% lower) but Comic Sans had only a 4.5% lower agreement rate than Baskerville, including their weighting system.

For Weighted Disagreement, Georgia had only a 7.7% increase in disagreement than Baskerville whereas the chart makes it look more than double.

Still interesting, but not nearly as substantial as they make it out to be. Is there a term for this type of manipulation of charts (whether intentional or not)?

EDIT: Indeed, the term for this is "Truncated graph" http://en.wikipedia.org/wiki/Misleading_graph#Truncated_grap...

And as a bonus (thanks wikipedia!), according to Edward Tufte's "Lie Factor"[0] (where 1 is considered accurate), the Weighted Agreement chart has a lie factor of ~15 and the Weighted Disagreement chart has a lie factor of ~17.

[0]: http://thedoublethink.com/2009/08/tufte%E2%80%99s-principles...

>Georgia is enough like Times to retain its academic feel, and is different enough to be something of a relief for the grader.

I've thought for years Georgia was a great choice on a resume/paper.

a) You want to stand out

b) You also don't want to appear too "starchy"

> Baskerville seems to be the king of fonts. What I did is I pushed and pulled at the data and threw nasty criteria at it. But it is clear in the data that Baskerville is different from the other fonts in terms of the response it is soliciting.

No amount of 'pushing' and 'pulling' at data can compensate for a poorly designed experiment. Georgia can't be used as both the control and a measure of how effective Georgia is—clearly fonts that stood out from the rest of the page would have a different effect than the one that looks exactly like the rest of the page. To give any of this credence, the sample should have stood alone, or the typeface of the surrounding page should have been randomized as well. What we're looking at here is "Are there certain typefaces that compel a belief that the sentences they are written in are true when contrasted with Georgia?".

Generally I detest font geeks, but I'm going to defend them here. First off: yes it's true that many intellectuals, designers, and hipsters have a genuine prejudice for Comic Sans. Facts need to work a little harder to prove themselves when written in that font. But in a world where we're deluged with typed information from the second we glimpse at our alarm clocks, I think it's okay to have a little prejudice because we need filter out at least some of the noise. Like most stereotypes, the Comic Sans prejudice is based on a grain (beach?) of truth. Can anyone really claim that the percentage of trustworthy Comic Sans based webpages they've seen in their life is equal to the percentage of trustworthy Georgia pages? Sorry. Geocities happened people, and I, for one, will never forget it.

This misses the most obvious difference to users, which is that the font changes in the middle of the article. Taken in conjunction with all the other fonts on the page, the harmony of the specific font to all other fonts in the article and on the page is probably the most important factor here.

I would like to ask Patio11 if he has ever done font A/B tests. I'm working on my sales website and the results would be very welcome. In my case I lack the traffic to do any proper testing.

Following up on blahedo's comment and the questions about what the heck their p-values mean --

This is a nice example that you can get statistical significance for small effects, if your sample is big enough. Their p-values are explained very badly, so I did my own analysis by transcribing their data from those plots. Let's take their weighting scheme for granted. I agree with some other commenters that the sums and counts are misleading, and instead took average scores per font, and computed confidence intervals for those means. The means are indeed a little different, and for some pairs, statistically significantly so.

http://brenocon.com/Screen%20shot%202012-08-10%20at%202.03.0...

But does it matter much? Take the pair with the largest gap, Baskerville vs. Comic Sans, of 0.95 versus 0.79: a difference is 0.16. This is out of a 10-point scale (ranging -5 to +5).

In fact, the standard deviation for the entire dataset is 3.6 -- so just 0.05 standard deviations worth of difference.

Or here's another way to think about it. If a person does Comic Sans example, versus could have done Baskerville example, how often would they have score higher? (This ignores the weightings, it's a purely ordinal comparison. I think this is related to the Wilcoxon-Mann-Whitney test statistic or something, I forget.) So with independence assumptions (if they had proper randomization, hopefully this solves), just independently sample from the distributions many times and compare pairs of simulated outcomes. 22% of the time it's a tie, 40.3% of the time Baskerville scores higher, and 37.8% of the time Comic Sans scores higher. I guess then it sounds like the difference is better than nothing.

Not sure what's a good and fair way to think about the substantive size of the effect. I wanted to take the quantile positions of the means, but realized you can't exactly do that with ordinal data like this (zillions of values share the same quantile value).

I probably missed something, so here's the transcribed data and R/Python code probably with errors: https://gist.github.com/3311340

Now that I'm thinking about it more, averaging the agreement scores seems weird. Maybe it's clearer to use the simple binary agree/disagree outcome.

I don't know about typefaces, but taking a screenshot of rendered black-on-white text and saving as .jpg sure has an effect on credibility.

I would like to see it controlled for age group, I remember liking Comic Sans as a child. Would Comic Sans have an effect on children, the same way it seems to for adults?

I'll take the other side here:

There's two axis - engagement and authority. Baskerville is not engaging, but it looks authoritative. So you tend to agree, even if you don't know what it says (like a boring professor or politician). Comic Sans is like a boring person in a clown suit - you can't follow what it's saying, and you tend to disagree just because it looks a little stupid.

The more respectabel Sans are engaging, but not authoritative; Times is both engaging and authoritative.

If you read something in Baskerville, you agree because it looks so boring that you can't be bothered reading it. Georgia, on the other hand, encourages both strong agreement and strong disagreement - people take it seriously, but actually pay attention. No-one takes Comic Sans seriously, because it's hard to read and looks stupid.

For those who don't have Baskerville: http://klepas.org/openbaskerville/ or http://openfontlibrary.org/font/open-baskerville/

"It's going to work! I'm using a very convincing font; it's bold, and has a lot of serifs."

http://www.youtube.com/watch?v=APcuJjCZTMU#t=4m07s

I find it odd that Comic Sans and Georgia change places in the weighted totals.

Now I'm going to petition Randall Munroe to gather more data (thinking of the color survey).

There are a lot of problems here.

The bar charts used to illustrate that article are terrible. They present raw counts for each font, but each font was not presented to the same number of people---they varied from 7,477 (CM) to 7,699 (Helvetica), which is a pretty big swing given the other numbers they're displaying. In fact, when you run the percentages, CM has a higher percentage of agreement than Baskerville (62.6% to 62.4%)!

When we turn to the "weighted" scores, which don't follow any clear statistical methodology that I'm aware of, the bar chart is again presented with counts rather than proportions, and this time with an egregiously misleading scale that makes it seem like CS gets half the score of gravitas-y fonts like CM and Baskerville, when in fact its score is only about 5% lower.

Finally we get to the "p-value for each font". That's... not how p-values work. The author admits that his next statement is "grossly oversimplified", but there's a difference between simplification and nonsense. He says that "the p-value for Baskerville is 0.0068." What does that mean? What test was being performed there? Can we have a little hint as to what the null and alternative hypotheses were?

Interesting read.

I certainly agree that Comic sans nudges me towards disbelief (and I'd never read a full article written in this horrible font :-)), while Georgia seems more 'professional' and believable.

Baskerville in my mind is instantly associated with all the books I read - most of those on scientific topics had this or a very similar font. Don't know whether it affects my judgement of what's written compared to any other normal fonts.

Typewriter-style fonts do make texts seem older and therefore, more believable (since they've been around for so long, there must be some truth to them - the standard logical reasoning).

So font has an effect on how seriously readers take what's written in it. The names of the fonts alone (i.e. Helvetica, Georgia, and Comic Sans) also give off the same vibes.

I wonder... Do the names of programming languages have an effect on how seriously people want to read what's written in them? If given 3 names (e.g. Python, Ruby, Groovy), do people subconciously rank their seriousness???

I took a class that was very writing-intensive. I was one of the only people to ever actually change the font in my essays from Calibri to Times New Roman, and always wondered if this contributed to the fact that I did substantially better than most other people with very comparable essays.

regarding comic sans use at CERN... often found that scientists like hideous designs. it's a way of saying that we work on serious stuff.

a lot of sites at MIT, CMU have that mark... the more prestigious, the uglier.

of course it has to be a certain style of ugly.

Our lecturer give us notes on particle physics in Comic Sans. I found them impossible to study from.

Edit: incidentally, she works at CERN on the ATLAS experiment too.

The ironic thing about this article is that it encodes the text images as JPEG. I wonder about image encoding's effect on credibility.

Honestly, the font had nothing to do with my decision - I just trust NASA.

As a scientist, I wouldn't worry about someone's lowered opinion who judges a scientific text by the font.