Problems with long-term monitoring of various extreme meteorological events (including tropical and extratropical cyclones, extreme winds, temperatures and precipitation, and mesoscale events) are examined. For many types of extreme events, the maintenance of long-term homogeneity of observations is more difficult than is the case for means of variables. In some cases, however, a strategy of using more than a single variable to define an event, along with the careful elimination of biases in the data, can provide quantitative information about trends. Special care needs to be taken with extreme events deduced from meteorological analyses, because changes in analysis and observation systems are certain to have affected extremes. Also, compositing of observations from more than one station, using differences in means (of temperature for instance) to produce a single long-term site, may not remove the biases in the extremes. These problems, along with ambiguities in defining extreme events, and difficulties in combining different analyses from different sites, complicate (and perhaps invalidate) attempts to determine whether extreme weather is becoming more frequent. The best that is likely to be achieved, even with increased emphasis on attaining the high-level of homogeneity necessary in the observations, is to monitor long-term variations in certain important extreme events, in select locations with high-quality data. Regional indices of important extreme events, selected on the basis of their damage potential and capable of adequate monitoring, may be established. If, in the future, we are to answer the question "Are extreme weather events becoming more frequent?", we must establish and protect high-quality stations capable of monitoring the most important extreme events (perhaps with such regional indices), and ensure that changes affecting the recording of extreme events (e.g., changes in exposure) are meticulously documented.