Inflation has an anchor in people's expectations of what its longrun value will be. If expectations persistently change, then the anchor is adrift; if they differ from the central bank's target, the anchor is lost. This paper uses data on expectations from market prices, from professional surveys, and from the cross-sectional distribution of household surveys to measure shifts in this anchor. The paper's main application is to the Great Inflation in the United States. The data suggest that the anchor started drifting as early as 1967 and that this could have been spotted well before policymakers noticed it. Other applications using expectations data from Brazil, Turkey, South Africa, the United States in the 1970s, and the United States in 2021 confirm the data's usefulness to measure the inflation anchor in real time.