This text is based on a Royal Society Meeting on Chaos and Forecasting held in London in 1994, that covered fields as diverse as biology, economics, epidemiology and earthquake science. It examines applications to nonlinear forecasting of both artificially-generated time series and real data.