Design has a citation problem.

A study gets published. A blog paraphrases it. Forty blogs paraphrase the blog. Nine years later it's a bullet point in a rebrand deck, presented as settled fact, while the original finding has failed to replicate three separate times and nobody went back to check.

Four of these are load-bearing in the way most people talk about typography. Two of them are on this site, in posts I wrote in May last year. Those posts now carry corrections, because the alternative is leaving them up and hoping nobody reads the source list.

Myth one: serif fonts are more readable

The claim: serifs guide the eye along the line, so serif faces are easier to read in long-form text. Sans is for screens, serif is for print.

The research says otherwise, and has for roughly ninety years. The modal finding across the literature is that there is no difference in the legibility of letters and words set in serif versus sans serif typefaces. Same result for comprehension of continuous text. Same result for overall reader preference. Alex Poole reviewed more than fifty studies and concluded the body of research was too weak and too inconsistent to support a claim either way. His summary: most reasonably designed typefaces in mainstream use will be equally legible.

What actually determines readability is spacing, size, contrast and resolution. All four of those vary as much within serif and sans as they do between them. Which means the category comparison was never the right comparison.

One exception survives, and it's narrow. Sans serif is slightly more legible in extreme conditions: text read at distance, at very small sizes, at low resolution, at speed, in poor light, or on safety signage.

Low vision is where people usually reach next, and it does not rescue the argument. The 2007 systematic review in the Journal of Visual Impairment and Blindness found the research had not produced consistent findings, and located no definitive evidence-informed standards for font legibility in low-vision materials. What the review did support was print size, contrast and spacing. The same variables. Again.

Neither of those describes your website body copy.

What to do with it: stop defending typeface decisions on readability grounds. The legibility difference is noise. The associative difference, what a serif signals versus what a geometric sans signals, is real, measurable and well documented. That's the argument. Make that one.

Myth two: hard-to-read fonts help people remember

In 2018, RMIT released Sans Forgetica. A deliberately awkward typeface, back-slanted with gaps in the letterforms, built on the theory of desirable difficulty. The premise was that making text harder to process forces deeper engagement, which improves recall. It got extraordinary coverage. Students installed it. Teachers printed with it.

In 2020, researchers at Waikato and Warwick ran four experiments with 882 participants. The paper is called Disfluent difficulties are not desirable difficulties, which tells you how it went. Sans Forgetica reliably produced a feeling of disfluency. It did not produce better recall. In some conditions it produced a memory cost.

Subsequent work replicated the null result repeatedly, including across free recall, cued recall and recognition testing. A meta-analysis of perceptual disfluency in study materials found no effect on later recall at all.

The genuinely useful part is why it spread. Sans Forgetica feels like it's working. Reading it is effortful, and effort feels like learning. It isn't. The sensation of difficulty and the fact of retention are two separate things, and people consistently mistake the first for evidence of the second.

What to do with it: friction that feels like effort is not the same as friction that produces effort. If your packaging is hard to read, it isn't making anyone concentrate. It's making them put it down.

Myth three: dyslexia fonts improve reading for dyslexic readers

Dyslexie and OpenDyslexic are both built on the same idea: weight the bottoms of the letterforms, exaggerate the differences between similar characters, and letters will stop flipping for dyslexic readers.

A 2017 study tested 170 children with dyslexia. They did not read Dyslexie faster or more accurately than Arial, and the majority preferred Arial. A separate study of OpenDyslexic found no improvement in reading rate or accuracy against Arial or Times New Roman, and not one participant reported preferring it. Another found participants strongly preferred Verdana and Helvetica over OpenDyslexic, with researchers recommending Helvetica, Courier, Arial, Verdana and Computer Modern on both performance and preference, and specifically cautioning against italics.

The reason is structural. Dyslexia is rooted in language processing rather than a visual defect. The fonts were solving the wrong problem, and solving it with real conviction.

The things that do help are unglamorous: size, spacing, contrast, line length. The same four variables that turned out to matter more than serifs in myth one.

What to do with it: if your accessibility statement leans on a specialist font, that's a decision you should be able to defend with something other than the font's own marketing. Researchers behind the 2017 study warned the fonts can offer false hope, which is a worse outcome than doing nothing.

Myth four: Baskerville is the typeface of truth

In 2012, Errol Morris ran a covert experiment through the New York Times. Readers were shown a passage in one of six typefaces and asked whether they agreed with it. Around 45,000 responses came back. Baskerville produced the highest agreement and the lowest disagreement. Comic Sans, predictably, came last.

Then the finding left the room without its context.

The effect was roughly 1.5 percent above average. Statistically significant given the sample size, and genuinely interesting. Also very small. Six typefaces is not a test of a category, it's a test of six typefaces, three of them serif and three sans, chosen without a systematic rationale. Morris said so himself: asked whether he'd proven anything he could stamp a Q.E.D. on, his answer was no.

What survives is the direction, not the magnitude. It sits comfortably alongside the broader fluency research showing that presentation format influences belief, including a study where statements shown at high visual contrast were judged true more often than the identical statements at low contrast. Format moves credibility. It moves it by a percentage point or two.

What to do with it: cite it as evidence that typography touches credibility. Do not cite it as evidence that a specific typeface makes people believe you. I had it in my own source list without the effect size for eight months. That's how this happens.

The shape all four share

A real finding, over-generalised, repeated until the caveat fell off.

And notice what keeps turning up once the myths are cleared out: spacing, size, contrast, weight, curvature. The unsexy variables that do most of the actual work, and which nobody writes a headline about, because "serif equals trust" is a much better sentence than "your line height is too tight."

The wrong version spreads because it's easier to say. That's not a typography problem. That's every problem in this industry.