There has been some research looking into whether reading something disfluent, like sans forgetica font, could improve learning via desirable difficulties. That is, when learning is more effortful, it often results in better long-term retention. Most recently, a meta-analysis (2) showed that disfluent font doesn’t lead to much better learning, but the hypothesis still might make sense for proofreading or revising. See, when font is disfluent you have to slow down to read it, which means that we can’t use those automatic processes that usually help us jump right over errors.
How did they test this?
At this point we have three groups: participants who are reading passages silently, aloud, or in disfluent font. While there were two experiments in the full study, I’m describing the second experiment here as the results are a bit easier to understand. While there are lots of different types of errors in real writing, the researchers narrowed in on errors that are less likely for a standard spelling and grammar check to find. These were some grammatical errors and some word choice errors.
Each text that was read had only one to five errors, so for the most part readers could read along without many errors.