In this paper, we reflect on the controversial concept of ‘rational expectations’ and point out that the meaning of ‘rational’ has changed over time. After briefly reviewing the literature, we describe the disaggregated firm-level data from the Bureau for Economic Research in South Africa. Our empirical investigation focuses on inflation expectations. The firm-level data are unique in breadth, scope and time span. We focus on these data, which are considered to be more representative of price setters than financial analysts or households. We compare these results with analysis of financial analysts, who have traditionally been the subject of these types of analyses and who tend to be relatively more rational. We find that while neither the inflation expectations of the business sector nor those of the financial analysts are rational in the strict sense of the term, both respond quickly to changes in underlying macroeconomic and financial conditions. There is evidence that inflation forecast errors stem partly from errors made in forecasting variables such as wages, the exchange rate and interest rates. We reach this conclusion because the survey data used also require respondents to forecast related economic variables. In addition, we find that important socio-economic factors such as firm size, the position of the respondent, and the industry a firm belongs to have significant effects on inflation expectations.