Testing of usability could perhaps be more accurately described as testing of learnability. We know more about the problems of novice users than we know of the problems of experienced users. To understand how these problems differ, and to understand how usability problems change as users change from novice to experienced, we conducted a longitudinal study of usability among middle-school teachers creating Web sites. The study looked at the use both the use of documentation and the underlying software, tracking the causes and extent of user frustration over eight weeks. We validated a categorization scheme for frustration episodes. We found that over the eight weeks the level of frustration dropped, the distribution of causes of frustration changed, and the users’ responses to frustration episodes changed. These results suggest that the sorts of errors that are most prominently featured in conventional usability testing are likely of little consequence over longer periods of time.