I found the vocab test itself [1] quite interesting, but without information about the distribution of scores it's difficult to interpret the "% words known" metric.
Is the distribution of scores publicly available anywhere?
All the words I did not recognize were either absolutely British regional slang or they had one or more synonyms that are overwhelmingly more common in the US.
Examples: juggins, freeview, jackfish, redbrick
The US equivalents would be bumpkin, superstation (?), northerns, and ivy (as in Ivy League).
I sort of assumed that the nonsense word generator would make nonsense compounds from valid English words, which is why I rejected those last 3, despite that nagging feeling that they were so simple that someone, somewhere might have used them.
I saw that when I clicked the "what does it mean" link at the end. I recall thinking that trademarks are not, strictly speaking, English words. That would be like saying "microsoft" is a word, because "Microsoft" is a company. That's why researchers in this vein have to review their source dictionaries, and scrub out all the things that met the lexicographers criteria but not the researchers'.
But maybe that is what they are doing. The test was on whether the viewer recognized the string as a word. The comparison at the end was probably just what can be found in the latest revision of the unabridged Oxford English Dictionary. They might be using the results to assemble a word corpus that could be used in other research applications. If that's the case, bravo for not requiring a bunch of research assistants to endlessly pore over word lists.
You're so afraid of the word "darkie" that you refuse to acknowledge (to an anonymous computer algorithm) that you recognize the word and close the window, yet you come to HN and tell the story using your real name?
Is the distribution of scores publicly available anywhere?
[1] http://vocabulary.ugent.be/wordtest/start