This paper explores the phenomenon referred to as test score inflation, which occurs when achievement gains on "high-stakes" exams outpace improvements on "low-stakes" tests. The first part of the paper documents the extent to which student performance trends on state assessments differ from those on the National Assessment of Educational Progress (NAEP). I find evidence of considerable test score inflation in several different states, including those with quite different state testing systems. The second part of the paper is a case study of Texas that uses detailed item-level data from the Texas Assessment of Academic Skills (TAAS) and the NAEP to explore why performance trends differed across these exams during the 1990s. I find that the differential improvement on the TAAS cannot be explained by several important differences across the exams (e.g., the NAEP includes open-response items, many NAEP multiple-choice items require/permit the use of calculators, rulers, protractors or other manipulative). I find that skill and format differences across exams explain the disproportionate improvement in the TAAS for fourth graders, although these differences cannot explain the time trends for eighth graders.