Just as a test, ticking all the boxes scores 45,000 words. Which seems to indicate that they haven't seeded the quiz with fake words to weed out cheaters : Pity, since there was an opportunity to unbias it in at least one dimension. (I also tried deselecting just 1 of a few of the really tough words : Each one caused the score to lower).
Site creator here -- you're right, there's no cheating detection. I ask people not to fill out the survey results if they're not being entirely truthful, of course they are free to disregard the instructions.
But I created the site, less interested in absolute vocabulary size numbers (these vary widely depending upon methodology), and more in relative changes among age groups, SAT scores, etc. And hopefully, cheating would not be correlated to any of those...
But at the end of the day, this is not a controlled, scientific survey. It is a voluntary quiz, though I am doing my best to control for other factors.
I'm really, really hoping that you're logging referrer data along with scores - if you're not already, you might consider starting. Even if this is not a source of proper scientific data, it would be highly interesting to see how, for instance, the Reddit referrals do vs. HN vs. Facebook, etc. We can probably all make some fairly accurate guesses about where the high and low performers would come from, but I'd be real curious to see the exact breakdowns.
