tl;dr — Fonts can influence the perceptions of the reader. Whether they can substantially raise your grades is an open question and no scientifically rigorous study seems to have been performed on the point. An anecdotal study on the point suggests that Georgia and Times New Roman will result in higher grades than Trebuchet MS. A more sophisticated study suggests that words set in Baskerville are more agreeable than words set in other typefaces.
You can read more about both studies in the NY Times article Hear, All Ye People; Hearken, O Earth by Errol Morris, which summarizes the anecdotal study on grades: "The Secret Lives of Fonts" by Phil Renaud:
Renaud had written 52 essays in total. Eleven were set in Times New Roman, 18 in Trebuchet MS, and the remaining 23 in Georgia. The Times New Roman papers earned an average grade of A-, but the Trebuchet papers could only muster a B-. And the Georgia essays? A solid A.
Errol goes on to note of the original article:
But Renaud’s results are anecdotal. I wondered: is there an experiment that could decide this once and for all? Or barring that, at least throw some empirical light on the situation? Could the effect of typography on the perception of truth be assessed objectively?
Here is how he performed a more trustworthy experiment:
Benjamin Berman (who designed the Multics emulation for my Times article “Did My Brother Invent Email with Tom Van Vleck?”) created a program that changes the typeface of the David Deutsch passage. Each Times participant read the passage in one of six randomly assigned typefaces — Baskerville, Computer Modern, Georgia, Helvetica, Comic Sans and Trebuchet. The questions, ostensibly about optimism or pessimism, provided data about the influence of typefaces on our beliefs.
The test consisted of comparing the responses and determining whether typeface choice influenced our perception of the truth of the passage.
More than 100,000 people clicked on the page, and approximately 45,000 people took the quiz.
The conclusions of the study being (my emphasis):
Baskerville is different from the rest. I’d call it a 1.5% advantage, in that that’s how much higher agreement is with it relative to the average of the other fonts. That advantage may seem small, but if that was a bump up in sales figures, many online companies would kill for it. The fact that font matters at all is a wonderment.
Here is a quote about the statistical significance of the Baskerville effect, from the article:
For those interested in the statistical analysis of the data, I have included a note from David Dunning. I had wondered whether the “Baskerville Effect” was so small as to be insignificant. A 1.5 percent advantage. Dunning assured me otherwise. He wrote to me, “In the 1990s, the federal government stopped a big trial testing whether taking aspirin prevented heart attacks. The aspirin worked, and it was considered unethical to prevent the control group from starting to take the drug. Size of the advantage to aspirin? It was 0.8%.”
David Dunning goes on to say, and I will quote at length from Part 2 of Errol Morris's article:
I have done three analyses, to see if they converge on the same conclusion. Two of them do; the third does not but involves a cruder and less sensitive measure, and so I am not too disappointed.
Scenario #1: The first analysis is to take the coding scheme you used to weight confidence (+3, +2, +1, -1, -2, -3) and ask this question. I redid the “omnibus” ANOVA that Ben carried out and got practically the same result, F (5, 45518) = 2.90, p = .013, any slight difference from Ben’s numbers are probably due to rounding error. What this result suggests is that the averages of the six font groups differ more from one another than one would expect by chance. Something somewhere is going on.
To get more specific, I next looked at each font in turn, to ask whether it produces responses that differ from the average of the other 5? When I do that analysis, these are the results I get:
Now, here are two catches. We never completely rule out random chance as a possible cause of any result we see. But, sometimes the result is so strong that chance is just very, very unlikely. What’s strong enough? If the p-value is .05 or less, we typically dismiss chance as an explanation, by “industry agreement”. That is, we tolerate a 5% chance on any one comparison that what we are looking at is merely random variation.
But here’s the second catch. I am doing 6 tests here, not 1 — and so there are 6, not 1, opportunities for me to be just looking at random chance. So I have to be more conservative in any call to dismiss chance as an issue. There are many ways to do this. For now, let me be just simple, blunt, and very conservative. I’m doing 6 tests and I want to give myself no more than a 5% chance of making an error, then let me simply divide .05 by the number of tests I’m doing. That means that the p-value needed for me to dismiss chance falls to .0083 (Or .83%). By this, Baskerville is still different from the average of the others.
If people prefer to ask instead to about differences between individual groups, I can confidently say people are responding to Baskerville more favorably than they are to Comic Sans (Tukey’s honest significant difference, q = 4.24). Given so many other group to group comparisons, it be-comes difficult to rule out chance for any other individual comparison.
Scenario #2: The coding scheme you used is reasonable, but it does have an unusual feature and there are alternatives. Specifically, there’s a gap right in the middle of your scheme between ‐1 and +1, and some might question whether having that gap is reasonable. That is, is moving from ‐1 (slightly disagree) to +1 (slightly agree) really twice as different as moving from +1 (slightly agree) to +2 (moderately agree)?
Interesting question, so I redid the analysis with a different coding scheme that removes this gap problem by stacking the responses up and placing the same interval in between them: -5(strongly disagree), -3(mod disagree), -1(slight disagree), +1(slight agree), +3(mod agree), +5(strong agree).
This is normally the way I would “weight” the confidence scores in my own research, so I’m comfortable with it.
The omnibus ANOVA calculation is significant, F(5, 45518) = 2.88, p = .013. Something’s going on.
Then I redid the analysis from comparing each font against the average of the others.
And, as you can see, this re-weighting scheme changes little. People are responding to Baskerville differently than they are the average of the other tests, even after correcting for the fact that I’m doing multiple tests. Any results for Comic Sans and Computer Modern weaken a touch; maybe, if generous, one could say that Georgia is “marginal,” but before correcting for chance. For individual group-to-group comparisons, the only one outstripping chance is the Baskerville to Comic Sans comparison (Tukey HSD q = 4.11).
Now, you may ask which of these coding schemes is the superior one. My answer is I don’t care. They are both reasonable. What is important is that they both lead roughly to the same conclusion.
Let’s just look at the data the most crude way possible, just counting up the percentage of times people agreed with Deutsch’s statement:
I switch from an Ftest to a ztest because the data are binary (i.e., agree, disagree) rather than more smoothly continuous. Here, no differences in fonts survive the chance correction put in place.
Essentially, what this is telling me is that throwing away the fine-grained information contained in the confidence ratings obscures the precision of the test.
But this analysis gives us a way to quantify the advantage to Baskerville. It’s small, but it’s about a 1% to 2% difference — 1.5% to be exact, which may seem small but to me is rather large. 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. Many online marketers would kill for a 2% advantage either in more clicks or more clicks leading to sales.
On the exact claim in question here, there is the study from the original article ("The Secret Lives of Fonts") that indicates that Times New Roman correlates with an A-, and Georgia with an A, significantly higher than the B- average with Trebuchet. However as Errol noted, it was an anecdotal study. That said, there is a correlation between fonts and the perception of the reader.
So the claim that Times New Roman (or Georgia) will lead to higher marks than Trebuchet (or other sans-serif fonts like Arial or Helvetica) seems plausible. It is plausible, in other words, based on the specific results of Phil Renaud's anecdotal study and based on the conceptual results of Errol Morris's study, that simply changing your font from a sans-serif to a serif (particularly Baskerville, but also potentially Georgia and Times New Roman) will lead to higher marks.
As an aside, I have written at some length about this as well.
1. Diemand-Yauman C, Oppenheimer DM, Vaughan EB (2011) Fortune favors the bold (and the italicized): Effects of disfluency on educational outcomes. Cognition118: 114-118 PubMed: 21040910[PubMed]
2. Kelley CM, Rhodes MG (2002) Making sense and nonsense of experience: Attributions in memory and judgment. In: Ross B, editor. Psychol Learn Motiv. New York: Academic Press; pp. 293-320
3. Zorzi M, Barbiero C, Facoetti A, Lonciari I, Carrozzi M et al. (2012) Extra-large letter spacing improves reading in dyslexia. Proc Natl Acad Sci USA, 109: 28: 11455-11459.10.1073/pnas.1205566109 PubMed: 22665803[PMC free article][PubMed]
4. Chall JS (1991) Stages of reading development. New York: McGraw-Hill
5. La Berge D, Samuels J (1974) Toward a theory of automatic information processing in reading. Cogn Psychol6:2: 293-323.10.1016/0010-0285(74)90015-2
6. Perfetti CA (1995) Cognitive research can inform reading education. J Read Res18: 106-115.10.1111/j.1467-9817.1995.tb00076.x
7. Katzir T, Wolf M, O’Brien B, Kennedy B, Lovett M et al. (2006) Reading fluency: The whole is more than the parts. Ann Dyslexia56:1: 51-82.10.1007/s11881-006-0003-5 PubMed: 17849208[PubMed]
8. Katzir T, Lesaux NK, Kim YS (2009) The role of reading self-concept and home literacy practices in fourth grade reading comprehension. Read Writ22:3: 261-276.10.1007/s11145-007-9112-8
9. Sweet AP, Snow CE (2002) Reconceptualizing reading comprehension. In: Block CC, Gambrell LB, Pressley M, editors. Improving reading comprehension instruction: Rethinking research, theory and classroom practice. San Francisco, CA: Jossey-Bass; pp. 17-53
10. Joshi RM, Williams KA, Wood JR (1998) Predicting reading comprehension from listening comprehension: Is this the answer to the IQ debate? In: Hulme C, Joshi RM, editors. Reading and spelling: Development and disorders. Mahwah NJ, USA: Lawrence Erlbaum Associates; pp. 319-327
11. Perfetti CA, Landi N, Oakhill JV (2005) The acquisition of reading comprehension skill. In: Snowling MJ, Hulme C, editors. The science of reading: A handbook. Oxford, England: Blackwell Publishing House; pp. 227-247
12. Cain K, Oakhill J, Bryant P (2004) Children’s reading comprehension ability: Concurrent prediction by working memory, verbal ability, and component skills. J Educ Psychol96:1: 31-42.10.1037/0022-0622.214.171.124
13. Conlon EG, Zimmer-Gembeck MJ, Creed PA, Tucker M (2006) Family history, self-perceptions, attitudes and cognitive abilities are associated with early adolescent reading skills. J Read Res29:1: 11-32.10.1111/j.1467-9817.2006.00290.x
14. Reading: RAND Reading Study Group for understanding: Toward an R&D program in reading comprehension.Washington, DC: RAND Education
15. Guthrie JT, Wigfield A (1999) How motivation fits into a science of reading. Sci Stud Reading3: 199-205.10.1207/s1532799xssr0303_1
16. Sweller J, Chandler P (1994) Why some material is difficult to learn. Cogn Instruct12:3: 185-233.10.1207/s1532690xci1203_1
17. Bjork RA (1994) Memory and metamemory considerations in the training of human beings. In: Metcalfe J, Shimamura A, editors. Metacognition: Knowing about knowing. Cambridge, MA: MIT Press; pp. 185-205
18. Craik F, Tulving E (1975) Depth of processing and the retention of words in episodic memory. J Exp Psychol Hum Learn104:3: 268-294
19. Bjork EL, Bjork RA (2011) Making things hard on yourself, but in a good way: Creating desirable difficulties to enhance learning. In: Gernsbacher MA, Pew RW, Hough LM, Pomerantz JR, editors. Psychology and the real world: Essays illustrating fundamental contributions to society. New York: Worth Publishers; pp. 56-64
20. Shea JB, Morgan RL (1979) Contextual interference effects on the acquisition, retention, and transfer of a motor skill. J Exp Psychol Learn 5:2: 179.10.1037/0278-73126.96.36.199
21. Castel AD, McCabe DP, Roediger HL III (2007) Illusions of competence and overestimation of associative memory for identical items: Evidence from judgments of learning. Psychon Bull Rev14: 107-111.10.3758/BF03194036 PubMed: 17546739[PubMed]
22. Alter AL, Oppenheimer DM, Epley N, Eyre RN (2007) Overcoming intuition: Metacognitive difficulty activates analytic reasoning. J Exp Psychol Hum Learn136:4: 569-576 PubMed: 17999571[PubMed]
23. Oppenheimer DM (2008) The secret life of fluency. Trends Cogn Sci12:6: 237-241.10.1016/j.tics.2008.02.014 PubMed: 18468944[PubMed]
24. Sungkhasettee VW, Friedman MC, Castel AD (2011) Memory and metamemory for inverted words: Illusions of competency and desirable difficulties. Psychon Bull Rev 18: 973-978.10.3758/s13423-011-0114-9 PubMed: 21626231[PubMed]
25. Song H, Schwarz N (2008) Fluency and the detection of misleading questions: Low processing fluency attenuates the Moses illusion. Soc Cogn26:6: 791-799.10.1521/soco.2008.26.6.791
26. Corley M, MacGregor LJ, Donaldson DI (2007) It’s the way that you, er, say it: Hesitations in speech effect language comprehension. Cognition105: 658-668.10.1016/j.cognition.2006.10.010 PubMed: 17173887[PubMed]
27. Rhodes MG, Castel AD (2009) Metacognitive illusions for auditory information: Effects on monitoring and control. Psychon Bull Rev16: 550-554.10.3758/PBR.16.3.550 PubMed: 19451383[PubMed]
28. Alter AL, Oppenheimer DM (2009) Suppressing Secrecy Through Metacognitive Ease Cognitive Fluency Encourages Self-Disclosure. Psychol Sci 20:11: 1414-1420.10.1111/j.1467-9280.2009.02461.x PubMed: 19845889[PubMed]
29. Lonsdale MD, Dyson MC, Reynolds L (2006) Reading in examination type situation- the effects of text layout on performance. J Read Res29:4: 433-453.10.1111/j.1467-9817.2006.00317.x
30. Reber R, Schwarz N (1999) Effects of perceptual fluency on judgments of truth. Conscious Cogn8: 338-342.10.1006/ccog.1999.0386 PubMed: 10487787[PubMed]
31. Werth L, Strack F (2003) An inferential approach to the knew-it-all-along phenomenon. Memory11: 411-419.10.1080/09658210244000586 PubMed: 14562871[PubMed]
32. Rhodes MG, Castel AD (2008) Memory predictions are influenced by perceptual information: Evidence for metacognitive illusions. J Exp Psychol Gen137: 615-625.10.1037/a0013684 PubMed: 18999356[PubMed]
33. McDaniel MA, Hines RJ, Guynn MJ (2002) When text difficulty benefits less-skilled readers. J Mem Lang46:3: 544-561.10.1006/jmla.2001.2819
34. Hughes LR, Wilkins AJ (2000) Typography in children’s reading schemes may be suboptimal – Evidence from measures of reading rate. J Read Res23:3: 314-324.10.1111/1467-9817.00126
35. Obrien B, Mansfiled JS, Legge G (2005) The effect of print size on reading speed in dyslexia. J Read Res28:3: 332-349.10.1111/j.1467-9817.2005.00273.x[PMC free article][PubMed]
36. Hughes LE, Wilkins AJ (2002) Reading at a distance: Implications for the design of text in children’s big books. Br J Educ Psychol72: 213-226.10.1348/000709902158856 PubMed: 12028609[PubMed]
37. Primor L, Pierce M, Katzir T (2011) Predicting reading comprehension of narrative and expository texts among Hebrew-speaking readers with and without a reading disability. Ann Dyslexia61:2: 242-268.10.1007/s11881-011-0059-8 PubMed: 21993604[PubMed]
38. Shany M, Lahman D, Shalem T, Bahat A, Zayger T (2006) Alef ad taf- A system for diagnosing disabilities in the processes of reading and writing according to national norms. Holon, Israel: Yesod Publishing
39. Dunn LM (1997) Peabody Picture Vocabulary Test. Minnesota, American Guidance Service.
40. Wolf M, Katzir-Cohen T (2001) Reading fluency and its intervention. Sci Stud Reading5:3: 211-239.10.1207/S1532799XSSR0503_2
41. Stecker SK, Roser NL, Martinez MG (1998) Understanding of oral reading fluency. In: Shanahan T, Rodriguez-Brown FV, editors. 47th yearbook of the National Reading Conference. Chicago: National Reading Conference; pp. 295-310
42. Pikulski JJ, Chard DJ (2005) Fluency: Bridge between decoding and comprehension. Reading Teach58:6: 510-519.10.1598/RT.58.6.2
43. Snellings P, van der Leij A, de Jong PF, Blok H (2009) Enhancing the reading fluency and comprehension of children with reading disabilities in an orthographically transparent language. J Learn Disabil42: 291-305.10.1177/0022219408331038 PubMed: 19223667[PubMed]
44. Jenkins R, Fuchs L, van den Broek P, Espin C, Deno SL (2003) Sources of individual differences in reading comprehension and reading fluency. J Educ Psychol, 95(4): 719-729.10.1037/0022-06188.8.131.529
45. Katzir T, Morris R, Lovett M, Wolf M (2008) Multiple pathways to dysfluent reading in subtypes of dyslexia. J Learn Disabil41:1: 47-66.10.1177/0022219407311325 PubMed: 18274503[PubMed]
46. Breznitz Z (2006) Fluency in reading: Synchronization of processes. Mahwah, NJ, USA: Lawrence Erlbaum Associates
47. Wolf M, Miller L, Donnelly K (2000) The Retrieval, Automaticity, Vocabulary Elaboration, Orthography (RAVE-O): A comprehensive fluency-based reading intervention program. J Learn Disabil33: 375–386.10.1177/002221940003300408 PubMed: 15493098[PubMed]
48. Yue CL, Castel AD, Bjork RA (2013) When disfluency is- and is not-a desirable difficulty: The influence of typeface clarity on metacognitive judgments and memory. Mem Cogn41: 229-241.10.3758/s13421-012-0255-8[PubMed]
49. French MMJ, Blood A, Bright ND, Futak D, Grohmann MJ et al. (2013) Changing Fonts in Education: How the Benefits Vary with Ability and Dyslexia. J Educ Res In press.
50. Nelson TO, Narens L (1990) Metamemory: A theoretical framework and new findings. In: Bower G, editor. The psychology of learning and motivation: Advances in research and theory. New York: Academic Press; pp. 125-173