“But when gathering information from the behavior of early adopters, there is an asymmetry in the importance we should attach to people getting things right vs people getting things wrong or being confused. Early adopters represent a sample of authors that is heavily biased toward the highly competent end of the spectrum. Out of necessity their pages are authored either based on reading the spec directly or on information that is only a few steps removed from the spec (by definition there are not many extant examples that they can copy through view source and the network of subtly misleading tutorials has not yet had time to get established). As such, early adopters using things as intended tells us very little; just that the features they are using can be used correctly by sufficiently competent authors. On the other hand, early adopters having difficulties is indicative of a feature that will be significantly more problematic when used by the (less well informed/ competent) population of authors as a whole. ”