@Iversen, Thanks for the reply. My intent in asking the question was the thought that perhaps your long-term attrition rate is not as high as you think, in that a previously forgotten word from a list may show up again in a subsequent list and then stick, though that would be semi-dependent on their frequency. Taking your words from various sources as you do, you have to come across words in some type of frequency, either overall for the language or in a topical domain. So perhaps the more frequent ones that are initially forgotten get listed again without an explicit plan for that to happen, while extremely infrequent ones do not. Thus perhaps your system "recaptures" a good part of words that are lost via attrition on the first go round. But this would be impacted by your subjective judgment in applying "guessable" that you discuss to whether to list a word or not.