The full-information rational expectations (FIRE) assumption is at the core of modern macroeconomics. We revisit recent evidence which rejects FIRE based on survey data. It relates forecast errors to news at different levels of aggregation. The evidence based on consensus forecasts testifies against the full-information assumption, the evidence based on data for individual forecasters against rational expectations. In contrast to earlier survey evidence that was largely dismissed as irrelevant, the recent evidence is likely to have a lasting impact for two reasons. First, the global financial crisis of 2007/08 has led to a certain uneasiness with the state of macro and a readiness to embrace new ideas. Second, the recent literature has put forward a number of promising alternative models of the expectation-formation process. We review these at the end of the paper.