Typography Is as Effective as Chocolate for Creativity

In 2006 a Microsoft Research and MIT team showed that twenty minutes of good typography produced the kind of mood-and-creativity uplift Alice Isen had recorded in 1987 by handing subjects a small bag of candy ^{1 2}. Four readers in the good-typography group solved a creative-insight task. Zero in the badly typeset group did ¹.

The font on your screen is doing more work than it advertises. The study is small. Microsoft Design itself published the line "Typography is as effective as chocolate at improving creativity" in 2024 ² — a popularization of the 2006 paper rather than a verbatim conclusion of it. The underlying claim is durable, replicated across six studies in the same paper: good typography measurably elevates mood, and elevated mood measurably improves performance on creative-insight tasks ¹. A reader at a beautifully typeset page does not just feel different. They think differently for the next problem they try to solve.

The study, briefly

Kevin Larson of Microsoft's Advanced Reading Technologies group and Rosalind Picard of MIT's Affective Computing group, with Richard Hazlett and Barbara Chaparro, ran six studies under the title "Measuring the Aesthetics of Reading" between 2005 and 2006 ^{1 3}. Two tested whether well-typeset text would produce the cognitive uplift Isen had produced with comedy films and candy in 1987 ⁴.

The setup was austere. Twenty participants per study, split into two groups of ten, were given a tablet device and twenty minutes with one of two versions of the same issue of The New Yorker ^{1 5}. One version was professionally typeset with optimal kerning, ligatures, and justified text. The other was set in Courier with words awkwardly spaced ⁵. After reading, participants were asked to solve Karl Duncker's 1945 candle task — affix a candle to a corkboard wall using only a box of tacks, a box of matches, and the candle itself, without dripping wax ^{1 4}.

Four of ten participants in the good-typography group solved it. Zero of nine in the bad-typography group did. (One participant in the poor-typography condition was discarded for prior familiarity with the task.) That is the centerpiece result — the one Microsoft Design retold for popular audiences as a chocolate comparison in 2024 ^{1 2 5}.

Larson also measured time perception (participants reading good typography underestimated the twenty minutes by an average of three minutes and eighteen seconds, indicating engagement; the bad-typography group underestimated by only twenty-four seconds) and corrugator-muscle activation, the eyebrow furrow associated with frowning ^{1 6}. Both pointed the same direction.

What they actually measured

The researchers measured three things, not one. Mood was inferred from time perception and from electromyography over the brow. Creative cognition was tested through the candle task and the Remote Associates Test, both standard instruments in the positive-affect literature ^{1 4}. Reading speed and comprehension were also measured, and crucially, did not differ between the two conditions ^{1 7}.

That last finding is the one most discussions of the study omit. Good typography did not make people read faster. It did not make them retain more facts. It made them happier and more willing to recombine ideas under time pressure — the kind of lateral thinking that solves the candle task, where the box holding the tacks must be re-conceived as a shelf for the candle. Isen's 1987 paper showed the same uplift in subjects who had just watched five minutes of bloopers or received a small chocolate bar; the typography paper showed that twenty minutes of well-set type does similar mood work ^{4 1}. Isen's positive-affect group solved the candle task at significantly higher rates than neutral controls ⁴. The typography effect is smaller in absolute terms but the direction is the same.

The paper's own framing is more careful than the chocolate headline. "Good typography induces a good mood," the authors wrote, "and when participants are asked to read text with either good or poor typography in two studies, the participants who received the good typography performed better on relative subjective duration and on certain cognitive tasks" ¹. There is no mention of chocolate in the paper itself.

Why this surprises

Most working assumptions about typography treat it as decoration. The font is what the page is wearing. The thinking is what the page is doing. The Larson study breaks that mental model in one direction: typography is not the dress over the body of the writing. It is part of the cognitive apparatus the reader is wearing while they read.

The surprise has two layers. The first is that something as quiet as kerning can produce a measurable shift in problem-solving on a corkboard-and-candle task. The second is that the shift looks like it travels through mood, not comprehension. The reader does not understand the page better. They feel better while reading it, and that feeling carries forward into the next problem they try to solve ^{1 4}. Typography is doing affective work, not informational work. It is closer to ambient music in a room than to the captions under a museum photograph.

It promises the mood in which the box stops looking like a box.

That is why the comparison to chocolate is more honest than it sounds. Isen's candy bar in 1987 did not teach subjects how to solve the candle task. It put them in the kind of mood in which the box stopped looking like a box ⁴. Twenty minutes of well-typeset New Yorker does, in this paper, something analogous ¹. The reading experience IS the thinking experience.

Why this matters for writers and designers

For writers, the typographic choices made by the platform you publish on are not neutral. A post set in default browser sans-serif at 14 pixels with cramped line height arrives as a different essay than the same words set in a 16-pixel humanist serif with a 1.6 line-height and a 65-character measure. The wrapper is part of the argument.

Robert Bringhurst's Elements of Typographic Style puts the band at "45 to 75 characters" for a single-column page, with sixty-six as ideal ⁸. Matthew Butterick's Practical Typography recommends "an average line length of 45–90 characters, including spaces" and 120 to 145 percent of point size for line spacing ^{9 10}.

For designers, small typographic choices in a product UI carry mood weight. Material Design 3 sets body text at 16 pixels and pegs the minimum body size at 16sp ^{11 12}. The WCAG 2.2 contrast minimum for body text is 4.5:1 ¹³. Those numbers are floors, not goals. Above the floor lies the territory the Larson paper hints at — type that lets the reader feel cared for, and therefore think.

For a note-taking app, the implication is structural. Notes are where readers do their thinking most often, and most never look at the typography of their own writing tool until something goes wrong. iA Writer's Oliver Reichenstein has argued that "what you do is designing text for optimal performance" and that "hardly anyone can discern good from bad typography, but everybody can feel it" ¹⁴. That is the Larson finding stated as a design philosophy.

What other research confirms or contradicts

The broader literature converges on two findings. First, line length matters more than font choice within reasonable ranges; the 45–75 character zone is consensus ^{8 9 15}. Mary Dyson's review at the University of Reading identified the 55 to 75 range as the preference zone, while screen reading tolerated longer lines up to roughly 100 characters without speed loss ¹⁵. Second, congruence between typeface and meaning measurably changes reading speed: the word "truck" is read faster set in Impact, "flower" faster in Snell Roundhand — the meaning of the word and the shape of the letters interact in the first few hundred milliseconds of recognition ¹⁶.

The counter-evidence is honest. The candle-task replication used twenty people per study, a sample size small enough that the four-of-ten result has wide confidence intervals ¹. The corrugator-muscle effect was statistically significant but modest ⁶. The reading-speed null result is unambiguous — good typography does not make you read faster ¹.

The skeptic's counter

A careful skeptic should note three things. The 2006 study tested twenty people. The condition contrast was deliberately extreme — professionally typeset New Yorker versus Courier with word-spaced runs — a gap larger than the gap between two reasonable real-world choices ^{1 5}. The follow-on effect on a single creative-insight task does not generalize automatically to longer-horizon writing, study, or decision-making.

The "as effective as chocolate" line is a popularization. It is accurate to the lineage — Larson's paper explicitly grounds itself in Isen's 1987 work, and Isen used chocolate as one mood induction ^{1 4} — but it is not a verbatim claim of the paper itself. The phrase originates in Microsoft Design's 2024 retelling by Rob McKaughan ². A careful writer should say: "good typography produced mood-and-creativity uplift on the same family of tasks Isen used with chocolate in 1987," not "good typography is literally as effective as chocolate."

What the study does not prove is that switching fonts will make you smarter. What it does support is that the affective wrapper around a piece of writing is part of the cognitive event the writing produces. Move the wrapper, and the event changes by a measurable but modest amount.

How to use this today

Treat typography as a quiet input to your own thinking, not a styling decision. Five concrete moves, free, all grounded in the consensus zone where Larson, Bringhurst, Butterick, Dyson, and Material Design agree ^{1 8 9 11 15}: 16-pixel body, 1.4–1.6 line height, 60–75 character measure, softened contrast, two typefaces. The principle: typography is upstream of thinking.

1. Set your body text at 16 pixels minimum, with line-height between 1.4 and 1.6, which is the 120–145 percent band Butterick recommends ¹⁰. Material Design 3 uses 16sp as the base body size ¹¹. Most default browser stylesheets ship at 16px; many themes drop it to 14px to fit more text on screen, which is a step in the wrong direction for the kind of reading that matters.

2. Aim for a measure of 60 to 75 characters per line. Bringhurst's ideal is 66 ⁸. Butterick's band is 45 to 90 ⁹. The Dyson literature places the preference zone around 55 to 75 ¹⁵. Anything wider exhausts the return sweep; anything narrower fragments the rhythm.

3. Soften the contrast. WCAG 2.2 sets 4.5:1 as the body-text floor ¹³. Pure-black-on-pure-white (21:1) is well above the floor and also above the threshold at which most readers feel comfortable for long sessions; Andrew Kim's "Minimally Minimal" set body text at #333 on #FFF, a softened contrast that reads as editorial rather than clinical ¹⁷.

4. Choose one serif and one sans-serif, then stop. Stephen Coles' Anatomy of Type documents 100 typefaces across seventeen classifications ¹⁹. You do not need seventeen. Most editorial design that reads as restrained pairs one humanist sans for body and one serif for headings, or the reverse, and lets weight and size do the rest of the work.

5. Read your own work in the typography you would want a stranger to read it in. The typography of your notes is the typography of your thinking ¹. If you take notes in a default sans at 12 pixels with 1.2 line height, you have agreed in advance to think in that texture.

6. In MNMNOTE, this is the default. Body text at 16 pixels with a 1.65 line-height, a 65-character editor measure, and a serif heading against a sans body — adapted from the original "Minimally Minimal" pairing with warmer text color (#1a1818) on white instead of cool gray on white ^{17 18}. The reading experience is the demo. A note app that ignores its own typography is asking the writer to think in a default nobody chose.

Frequently Asked Questions

Did Microsoft researchers really say good typography is as effective as chocolate?

The paper itself, "Measuring the Aesthetics of Reading" (Larson, Hazlett, Chaparro and Picard, 2006), does not contain the word chocolate ¹. The chocolate framing comes from Microsoft Design's 2024 popularization by Rob McKaughan, which connects the study to Alice Isen's 1987 paper that used a small bag of candy as one of several mood inductions ^{2 4}. The lineage is real. The phrasing is a retelling.

What was the actual finding of the Larson and Picard study?

Across six studies, the researchers showed that well-typeset text produced measurable mood uplift (via time perception and corrugator EMG) and improved performance on creative-insight tasks. In the candle-task study, four of ten good-typography readers solved it; zero of nine bad-typography readers did ^{1 5}. Reading speed and comprehension did not differ between conditions ¹.

What is the candle task?

Karl Duncker's 1945 problem: given a candle, a box of tacks, and a box of matches, affix the candle to a corkboard wall so it burns without dripping wax on the floor. The solution requires re-conceiving the tack box as a shelf — the cognitive move called functional fixedness, which is the move positive-affect interventions help subjects make ⁴.

What font and layout did Microsoft use for the "good" typography condition?

The participants read a specially typeset issue of The New Yorker on a tablet device, with proper kerning, ligatures, and justified text ^{1 5}. The "bad" condition was set in Courier with awkward word spacing. The paper does not name a single specific typeface for the good condition because the contrast was the full set of professional typesetting choices, not one font ¹.

Has anything changed in default fonts since the study?

In July 2023 Microsoft announced Aptos, designed by Steve Matteson under the working name Bierstadt, as the new default font for Word, Excel, PowerPoint and Outlook, ending Calibri's seventeen-year run as the Microsoft 365 default since 2007 ^{20 21 22}. The rollout reached most general users between mid-December 2023 and late January 2024 ²³. The change became a TikTok story within forty-eight hours of the rollout — evidence that everyone, even people who claim not to care, has font opinions ²⁴.

Is there a single line-length number I should remember?

Sixty-six characters per line, including spaces, for a single-column page. The band from forty-five to seventy-five characters is the consensus zone across Bringhurst, Butterick, Dyson, and Schriver ^{8 9 15 25}. For multi-column work, a lower band of forty to fifty characters per column ⁸.

Does the chocolate effect mean I should ignore content and just style my notes prettier?

No. The Larson study did not show that typography substitutes for thinking. It showed that twenty minutes of well-typeset reading produced about as much mood-and-creativity uplift as Isen's chocolate-bar mood induction in 1987 ^{1 4}. Substance is upstream of typography; typography is downstream of substance and upstream of the reader's next problem.

What the world looks like if this is true

The font on your screen is not a costume. It is part of the room your thinking happens in. A small Microsoft Research paper from 2006 suggests that twenty minutes of well-set type can carry forward into the next problem you try to solve ¹. The effect is small. The effect is real.

If your notes are where you think, the typography of your notes is the typography of your thinking. A page set with care does not promise smarter ideas. It promises the mood in which the box stops looking like a box.

Typography is not decoration. It is how you think.

References

Academic / primary research

Mainstream press / authoritative publications

Industry / standards bodies

Personal sites / editorial / expert voices

Community / popularization / supporting

Standards / specs