A one-time survey may give a falsely precise indication of local usage. Examining four iterations of a library assignment survey reveals large within-discipline variation; even individual faculty members are inconsistent in their use of library assignments from year to year. Additional causes of variation include changing faculty and pedagogy. This article examines data from a survey sent to faculty about library assignments in their courses in 1996-1997, 2001-2002, and 2003-2004 at a small private masters-level college, and in 2004-2005 at a large public doctoral-intensive university. The researcher expected to discover how coursework in different disciplines required different levels of independent information seeking ("library usage"). The survey method was chosen when the contents and formats of course syllabi proved too inconsistent to yield the needed information about usage of library assignments. Use of library assignments was expected to be relatively consistent from year to year, and from institution to institution, because of the assumption that discipline strongly affects use of library information sources. Each time, the survey achieved a good response rate and gave apparently valuable information about current library assignments. However, the expected disciplinary consistency was much less than anticipated. The variation from year to year within disciplines--an average of sixteen percentage points--was almost as great as the variation between disciplines in any one year--an average of 18 to 29 percent. This article describes the intent, scope, focus, and initial findings of the original surveys, then uses the data from the four together to explore potential causes of the year to year variation. The results of this secondary analysis suggest that faculty use of information-seeking assignments is much more volatile than any one-time survey might show.