In this commentary stimulated by Fritsche et al.’s (2014) paper on ‘‘Forecasting the Brazilian Real and Mexican Peso’’ and the implications for forecast rationality, I first survey the literature on forecaster behaviour, and conclude that organisational and psychological factors heavily influence the characteristics of the forecasters’ errors in any particular application. Econometric models cannot decompose the error into these potential sources, due to their reliance on non-experimental data. An interdisciplinary research strategy of triangulation is needed if we are to improve both our understanding of forecaster behaviour and the value of such forecasts.