- No Recommendations yet - go add some!
- At first I agreed with those suggesting that Jonah, Jad, and Robert cherrypicked instances where re-tests exhibited steady declines, but if we assume that each experiment was done properly (as they noted, Schooler seems to be regarded as being fundamentally sound in his methods) and if both the participants and research assistants varied randomly in re-tests (ruling out unconscious bias or habituation on either of their parts), then it would seem surprising to see this even once, let alone several times in different domains?
Secondly, test-retest is already a kind of meta-analysis, so you're essentially saying that we need a meta-analysis of these meta-analyses, at which point the true mean and the true random distribution around it will reveal itself. But why not "zoom out" a level further and do a meta analysis of the meta analyses of the meta analyses? I'm not trying to be cute; only pointing out what I perceive to be a flaw in the logic. At which point would we expect to no longer see the decline effect, or be satisfied that we happened upon it by chance and that its not indicative of something real?
