This paper addresses two audiences--those who design education trend studies that simultaneously have longitudinal and intercohort implications, and the secondary analysts who use such trend data. Four study series conducted for the U.S. Department of Education's National Center for Education Statistics (NCES) provide material for the paper: the National Longitudinal Study of the High School Class of 1972 (NLS-72), High School and Beyond (HS&B), the National Education Longitudinal Study of 1988 (NELS:88), and the Education Longitudinal Study of 2002 (ELS:2002). The paper considers the balance between the need for continuity--to ensure comparability of findings and estimates over time; and change--to adapt to new educational circumstances and evolving policy and research agendas, and exploit new technological opportunities for improving data collection, data quality, and data accessibility. For the benefit of analysts of the four data sets, intercohort comparison points are identified, and caveats about aspects of non-comparability (ranging from changing eligibility and exclusion criteria to mode of administration effects) entered. For the benefit of survey developers, the following recommendations are made: definitions and procedures must be standardized to the extent possible; departures from past practice must be thoroughly documented; more methodological work (including "bridge" studies) is needed to test the implications of changes in content, format, and procedure over time. (Contains 1 figure, 19 tables, and 35 footnotes.)