Slashdot Mirror
Baskerville Is the Greatest Font, Statistically, Says Filmmaker Errol Morris
An anonymous reader writes "A survey of unsuspecting New York Times readers implicitly answered the question: Does a certain font make you agree or disagree more often than another font? It turns out Baskerville confers a 1.5% advantage towards agreement on a survey question, compared to an average of six fonts. They were asked to agree or disagree to a passage from physicist David Deutsch's book The Beginning of Infinity, and were found to have an optimistic, if Baskerville-favoring, outlook on life. David Dunning, a psychologist awarded a Nobel prize and, separately, an IgNobel prize (for the eponymous Dunning-Kruger Effect), called Baskerville 'the king of fonts.' Sadly, Comic Sans — notable for its appearance in the Higgs Boson announcement — seems to be the weakest font. And why did Lisa Randall, the Harvard physicist responsible for that Higgs announcement use Comic Sans? According to the article, 'Because I like it.'"
http://www.youtube.com/watch?v=wkoX0pEwSCw
"Comic Sans - notable for its appearance in the Higgs Boson announcement" "Because I like it."
Are you trying to kill the weak among the font aficionados? Comic Sans may be notable, but not for anything good.
Baskerville: http://en.wikipedia.org/wiki/Baskerville
Open Baskerville: http://klepas.org/openbaskerville/
And why did Lisa Randall, the Harvard physicist responsible for that Higgs announcement use Comic Sans? According to the article, 'Because I like it.'"
Given the mostly fixed number of neurons available to any single individual, the talent for physics must have come from somewhere... obviously, the aesthetics circuits got the short end of the deal.
Ezekiel 23:20
And..
Statistics are like bikinis, what they reveal is suggestive but what they conceal is vital.
and Lisa Randall was not the responsible for the announcement.
Lisa Randall, a Harvard physicist, kindly e-mailed Fabiola Gianotti on my behalf. Gianotti, the coordinator of the CERN program to find the Higgs boson, provided a compelling rationale for why she had used Comic Sans. When asked, she said, “Because I like it.”
Lisa *asked* the responsible.
Oh editors, I miss the times where at least you read the submitted articles. Now the anonymous guy can write whatever he wants in the summary and you'll publish it.
Where are the error bars?
Achewood:
http://achewood.com/index.php?date=07052007
--
BMO
Almost everyone has a reaction to it, positive or negative. Few people see it and just read it.
Damn_registrars has no butt-hole. Damn_registrars has no use for a butt-hole.
I think that an interesting follow-up study would compare subject matter and typeface pairing. That is, I believe that an article in physics is more likely to be taken seriously if it is set in a typeface (not a font, btw) like Baskerville than in comic sans. But what if you're subject matter is meant to be humorous? I suspect that people find it funnier if it is written in the comic sans than if it is written in Baskerville. Also, what typeface are people accustomed to reading such material in? Experience may play a large factor.
Anyway, it’s an interesting result, all the same. I'm sure the marketers will be thrilled to discover that they could grab another 1.5% if they'd just use the proper type.
The summary misstates the person responsible for using Comic Sans in the Higgs boson announcement. The full quote:
Lisa Randall, a Harvard physicist, kindly e-mailed Fabiola Gianotti on my behalf. Gianotti, the coordinator of the CERN program to find the Higgs boson, provided a compelling rationale for why she had used Comic Sans. When asked, she said, “Because I like it.”
I was already wondering why a Harvard physicist would be making the announcement of a discovery by CERN.
Time to update my resume.
This is probably a result of an occupational hazard, but I know very well that I pay more attention to text typeset in Computer Modern. Even though it is the default font in LaTeX, to that what Times New Roman is to Word.
xkcd is not in the sudoers file. This incident will be reported.
So, basically, world history might have taken a different turn if Al Gore's campaign had used Baskerville. And wouldn't Comic Sans have been the perfect match for 43? Ah, democracy, lead us onwards.
I'm really fond of the Dunning-Kruger effect to the point where I mention it almost daily and people get annoyed with me. So I was really surprised to hear the claim in summary that Dunning had a Nobel. What would it be in? The last time a psychologist got a Nobel it was for work related to economics. Sure, enough 10 seconds of fact checking, verified that he's not on any list of Nobel Laureates, such as http://en.wikipedia.org/wiki/List_of_Nobel_laureates or the official lists at Nobelprize.org. The claim about Dunning getting a Nobel isn't in TFA so I'm not sure where it came from.
You should send 10M€ to my bank account.
It's not a bad, font... it's actually a great font. It's people using it in appropriately that bothers you.
In the early 80's everyone with Print Shop was obsessed with outline fonts and old west fonts
In the Early 90's it was all the Calligraphy fonts.
All of them have their use. Some more general purpose than others.
Anyone who gets their panties twisted up over use of Comic Sans for its own sake is an asshat
I'm not sure I believe this story. But I probably would if it were displayed in Baskerville.
A serif font? Fail.
Calibri and Arial are miles more readable.
Its really depressing that with all the new fonts, studies of perception, cognition, etc, the greatest font is one that was designed in 1757.
Dyslexia
While not the best, it's decent and by far the most widely available (of the fonts dyslexics find easier to read).
compared to an average of six fonts
Err, what? I think it was compared to exactly five other fonts.
systemd is Roko's Basilisk.
"It would be cruel to say that [people who complain about the use of Comic Sans on the Higgs Bosson announcement] picked on the font because it was the only part of the presentation they understood [...but...] that's exactly what happened. The scientists had just used a machine that makes the Saturn V moon rocket look like a sparkler to interrogate reality itself, and these dumbasses were trying to look superior because they prefer letters with curly bits at the ends."
source
Just plain courier, it was good enough for gramps, it's good enough for me too. No need worry about your fancy proportional fonts like you're king of France or something.
Do the 'scientists' at NYT know that it's highly possible that a lot of Windows users might not have the font installed on their system AT ALL?
Just because they use Mac, they assume everybody MUST use mac.
I found the article very interesting, but can't get over the fact that they had no control on what the end user was seeing. And a LOT of them were likely seeing Times New Roman instead of the mac font.
"Yo, Baskerville, I'm really happy for you, and Imma let you finish, but I just got to say Helvetica is one of the best fonts of all time."
So Errol Morris is the guy responsible for the greatest advertising campaign of all time. http://www.errolmorris.com/commercials/miller.html
I won't say High Life is one of the beers I prefer because of these advertisements, but they sure are a part of why I first tried it. (its a favorite of mine because 6-pack of bottles for less than 5 dollars of a beer that is extremely drinkable is most definitely gold.)
I'm a french guy, and I never saw the Baskerville font used in France.
I'm pretty sure that this font has a cultural connotation for english people, but not for the rest of the world.
When one reads a text in Baskerville, one probably unconsciously associates it with ancient books, and with ancient wisdom.
An interesting experience would be to write a "modern" question (using recent words) with Baskerville, and measure its impact.
Baskerville looks like the font used by 1800's snake oil salesman.
In the Typography course I took, we were taught that the greatest font of all time is Garamond.
It wasn't even tested in this article.
Life is like a web application. Sometime you need cookies just to get by.
No, not that one. Or this one.
When I started at a NASA center, working with a bunch of physicists for the most part, I found I was being sent to an AAS (American Astronomical Society) meeting. I don't remember exactly what my boss said that was disparaging about astronomers, but I do remember he said something to the effect, 'but at least they're not mathmeticians, as they generally bathe at least once a week'.
So, just remember -- they might've been cleaning themselves out of the sink -- but at least they were cleaning themselves.
(and well, during undergrad, I think I had a period of about 10-14 days when I don't think I went above ground ... at least not when the sun was out (and it was summer) ... the problem is, you can't tell just how ripe you've managed to get ... so engineers aren't always the best group, either).
Build it, and they will come^Hplain.
Determining there is a difference between two things because one is significantly different than a reference and the other is not.
He also doesn't say what was compared. And the result is pretty marginal. Interesting, but definitely not the law of nature he implies.
How would you know, when I viewed the linked quiz it was helvetica. The 1.5% might be linux users!
They responded to a passage from asked to agree or disagree to a passage
Good thing the editors were not involved in this study. They would have had to read the passage... and then respond to a passage from asked to agree or disagree to a passage.
I mean, honestly.
...result from a poorly-designed survey. This is the kind of stupefying trivia that is the hobgoblin of little minds, bad designers and their far too numerous acolytes. cf. "Helvetica", the movie.
It is Fabiola Gianotti that used Comic Sans because she liked it. Not Lisa Randall. RTFA - or at least the relevant parts.
http://www.acetonestudio.com
...I might get some good karma!
The CB App. What's your 20?
While it is the rare occasion in Slashdot when we are talking about fonts, I gotta mention that I have become really enamored to Segoe UI as a screen font. So this is the default UI font you see in Windows 6.x. I use it even in Ubuntu, heh.
the summary is written in another font, so i must disagree.
Ok, because of that and that the choosing on the font probably is influenced by other things, popular enough (don't know, i.e. old prints of the bible or old style scientific papers) written in that font or similar enough ones in key aspects that rigs our judgement.
When I bought Kindle one of the things I wondered is why does this thing allow me change the font? Isn't it supposed to be the part of the book composition?
Dunno if it helps or hurts in non-fiction books, but IMO it really helps the author to convey the mood in fiction literature.
May Peace Prevail On Earth
FTA -- "Are the results the product of chance? To address this question, Dunning calculated the p-value for each font. Grossly simplified, the p-value is an assessment of the likelihood that the particular effect we are looking at (e.g., the effect produced by Baskerville) is a result of a meaningless coincidence. [10] The p-value for Baskerville is 0.0068 [snip explanation of P-values]... The conservative approach is to divide 5 percent by the number of tests. Thus, the p-value to dismiss chance falls to 0.0083."
While being a bit unclear on what's being tested or how (no stated null hypothesis, etc.; not surprising for the mass-media New York Times), the thing that's really sketchy is how this comes after this arbitrary weighting of "strong agree" as +5, "strong disagree" as -5, etc., generating single scores for each font which are wildly more divergent that the initial raw agreement bar-charts. I suspect that if P-values are allegedly being computed for these decorated scores, then the results would be invalid (I'd love to know what statistical test he thinks can be run on arbitrarily mangled scores like that).
Let's run a two-proportions z-test on this data -- I'm looking at Weiss Introductory Statistics, Procedure 12.3, but you can also see the procedure here. The hypotheses are Ho: Baskerville has the same proportion in agreement as other fonts, versus Ha: Baskerville has a different proportion in agreement from other fonts; this is a two-tailed test, since we didn't know in advance whether it should be higher or lower. For Baskerville x1=4703 (number agreeing), n1 = 7536 (total number surveyed), so p1^ = 0.624 (sample proportion agreeing); for other fonts x2 = 23265, n2 = 37988, p2^ = 0.612. The pooled proportion is pp^ = (x1+x2)/(n1+n2) = 0.614. The standard error is SEE = sqrt(pp^(1-pp^)*(1/n1 + 1/n2)) = 0.00793. So the test-statistic z-score is z = (p1-p2)/SE = 1.47. And the P-value for a two-tailed test is P = 2*F(-|z|) = 2(0.0708) = 0.1416 (this last from a table lookup).
In conclusion: While the article claims a P-value of 0.0068, and that we should require a P-value of less than 0.0083 to indicate strong evidence for the article's hypothesis, when I do a simple proportion test, the P-value is actually much higher: P = 0.1416. In my book this is interpreted as "weak or no evidence" -- obviously much higher than the customary 5% cutoff, even ignoring the need to divide by the number of tests. So it appears that we do not have statistically significant evidence for the article's findings.
We know where leadership by an anti-intellectual "strongman" who scapegoats minorities and likes boisterous rallies goes
I'm no fontographer, but I really like Gentium, and have for many years. It's capitals are not as tall as the ascenders, and the bow of h and n give a distinctive flavor to the text. With Gentium the two letters r and n run together (rn) look nothing like the letter m, but with Baskerville it is difficult to distinguish them.
Any other takers for Gentium?
--Udo.
This study just corroborates what I already knew. People are shallow. Especially all you Comic Sans haters. I think half the Comic Sans hatred is just because it has become trendy to hate Comic Sans.
I wanted to confirm the statistical analysis with a Monte Carlo simulation that assumed that all fonts had similar outcomes (null hypothesis). I ran 22 simulations to examine the range of possible "true" values of the six categories of choices. Each simulation used 45.5k answers... needless to say there is good confidence in the average number of people who chose each category but the error should not be ignored. For example, the model predicted a confidence interval of "strong supporters" ranging between 20.6 and 21.3% (vs. observed 21% in study).
These values obtained from the first set of simulations were used in a subsequent simulation of apparent choices of 7500 people. For each of the 22 subsets of runs, I ran 36 Monte Carlo simulations, leading to 792 results. The predicted values of the net supporter score (using the calculation proposed by the author) ranged between 0.64 and 0.95. Nine of the values were above the value observed for Baskerville (0.885). 9/762*6 = 7% chance that the observed conclusion of significance is true (1-sided test).
Perhaps I need to run more simulations to narrow down the confidence interval of my conclusions, but I certainly don't have enough certainty to state that the results are due to true differences between the fonts; I am sure that if you throw out the ridiculous inclusion of Comic Sans from the list of fonts, the authors would conclude that there was no significant differences between the other fonts. I personally would think that the world is ending if every quiz was written in Comic font!
Baskerville is a good font. Maybe not the best for everything, but nice. Definitely preferable to shit like Bleeding Cowboy
Back when I was a scriptwriter for a university theatre group, I did a small test with a few fonts: Palatino, Times, New Century Schoolbook and maybe a few others. Nothing scientific, just me and a couple of friends looking at small passages for readability etc. Palatino was the clear winner there, for readability and for general artistic atmosphere.
Later I came across an older script (written before my time) and lo and behold, it was written in Palatino.
(... and the movie scripts are written in Courier. For the life of me I cannot fathom that)
I hope part 2 of this insightful piece talks more about the role that context plays in our attitudes toward typography. Try putting Baskerville on a street sign and see how that plays out.
I believe that, in spite of the huge sample size, the claimed effect of Baskerville is not statistically significant.
Please find my statistical analysis in R code here: http://pastebin.com/WjDdiS8T
If there is any effect, it looks to me like it is only from Comic Sans giving 1.2% less "yes". And perhaps there is some other explanation for this, e.g. a small (1%) group of people breaking off because of seeing Comic Sans.