Processing Difficulty Effect: What Hard-to-Read Text Does to Memory

The effect, where it appears, is explained as a desirable difficulty: a feature of the material that makes learning feel harder while making it more durable. Robert Bjork introduced that term for practices such as spacing study sessions, testing yourself and generating answers rather than reading them, all of which have strong records. The proposal behind the processing difficulty effect was that an awkward typeface could join the list, because the feeling of effort it produces would prompt deeper, more careful processing of the words, and deeper processing is remembered better.

The replication record says otherwise for typefaces in particular. Rummer, Schweppe and Schwede pointed out in 2016 that the hard-to-read materials in the original experiments were not only harder to read but unusual, and so might have drawn attention by being distinctive rather than difficult. Their three experiments, close replications of the original first study, found no effect of the font on learning. Xie, Zhou and Liu's 2018 meta-analysis of text-based studies found pooled effects of disfluency on recall and on transfer close to zero. Weissgerber, Brunmair and Rummer reanalyzed that meta-analysis in 2021 and found its transfer result robust, while raising doubts about the recall analysis and the moderators, so the fairest summary is that no reliable benefit of hard-to-read type has been established, and a few boundary conditions remain in dispute.

One difficulty has a better record than typeface. Sungkhasettee, Friedman and Castel found in 2011 that words printed upside down were recalled better than upright words, in a single list and across repeated study-test cycles. The same study found that people's predictions of their own recall did not distinguish the two formats at all, which is the second and sturdier half of this topic.

Predictions of learning follow fluency. Rhodes and Castel showed in 2008 that words in a larger font were given higher judgments of learning although font size made little difference to recall, and that the effect persisted through warnings and repeated tests. Yue, Castel and Bjork found in 2013 that clear words were given higher judgments of learning than blurred words in every experiment that mixed the two formats in one list, by about half a standard deviation in the first, and the blurred words never did better: recall favored the clear words marginally at half-second exposures and significantly at two-second exposures, and did not differ when the words were tested by recognition or shown for five seconds. The feeling that clean text is sinking in is a judgment about reading, and it is a poor guide to what will be remembered.

Connor Diemand-Yauman, Daniel Oppenheimer and Erikka Vaughan published "Fortune favors the bold (and the italicized): Effects of disfluency on educational outcomes" in Cognition in 2011. Their starting point was earlier work, including Oppenheimer's own, showing that the subjective experience of difficulty leads people to process material more carefully. If so, making learning materials harder to read might improve memory for them.

Study 1 was a laboratory experiment. Participants studied information printed either in an easy-to-read font or in harder-to-read fonts, and were later tested on it. Those who had read the hard-to-read material remembered more of it. Study 2 moved the manipulation into high school classrooms, where class materials were reproduced in harder-to-read fonts for some classes and left as they were for others, and the students who received the harder fonts did better on their assessments. The authors concluded that superficial changes to learning materials could yield significant improvements in educational outcomes, and the paper was widely reported.

Two features of the paper matter for reading the record that followed. The hard-to-read fonts were also unusual, so difficulty and distinctiveness were confounded, as Rummer, Schweppe and Schwede noted when they could not reproduce the result. The claim also traveled further than the data: a laboratory list and one school's classes became advice to print study material in awkward fonts. A commercial typeface designed on that advice, Sans Forgetica, was tested in its turn and did not improve memory either. The original authors were careful about caveats; the finding's popularity was not.

The rows below are illustrations of where reading ease gets mistaken for learning, and of advice that leans on the original finding. They are not findings from the studies cited on this page.

There are two things to counter here: the advice to engineer difficulty with typefaces, and the fluency judgment that gives the advice its appeal.

• Do not print study material in hard-to-read fonts expecting a memory gain. The original result has not replicated reliably and the meta-analytic estimate is close to zero.
• Spend the effort on difficulties with a record: test yourself instead of rereading, space your study sessions, and generate answers before you look them up. These are the desirable difficulties Bjork had in mind.
• Treat the feeling of smooth reading as information about the text, not about your memory. Rhodes and Castel's participants kept rating large-font words as more memorable through explicit warnings; the judgment is hard to switch off, so check it with a test rather than arguing with it.
• When a text feels hard, slow down and summarize it from memory. The difficulty tells you the text needs work; it does not guarantee the work has been done.
• When you evaluate teaching, a lesson or a course, ask for a delayed test rather than a rating of how clear it felt.

The processing difficulty effect is one proposed desirable difficulty among several, and it is the one with the weakest record. Its neighbors are easy to confuse with it.

• It is not the generation effect, where producing information yourself, such as completing a word from a fragment, improves memory for it. Generation is a difficulty in the work of learning; an awkward typeface is a difficulty in reading.
• It is not the testing effect, where retrieving material on a test strengthens it more than restudying does. Testing has a large and reliable literature behind it.
• It is not the Von Restorff effect, where one item that stands out from the rest is remembered better. Rummer and colleagues argued that the original fonts may have worked this way, by being unusual rather than difficult, and that a distinctiveness effect cannot survive making every item distinctive.
• It is not the levels of processing effect, where attending to a word's meaning rather than its appearance improves recall. That effect concerns what you do with the material, not how hard it is to see.