Recently, there has been increasing concern about indicators for quality of care systems. Much energy has been devoted to the development of these indicators, but after a couple of years many professionals in the care organizations are complaining that the validity is extremely doubtful. In this editorial a fundamental problem is discussed, namely the reliability of the observations. In all scientific fields many precautions are necessary, for example in meteorology measuring the temperature in open air. In the manuals about the indicators of care two methods of observation are mentioned. In the first one patients are interviewed. The results seem very unreliable, because of socially desirable answers. For example, in nursing homes the patients are extremely dependent on the care-giver and it is likely they avoid complaints. The second method is used for gathering data about the frequency of falls, decubitus and so on. This is carried out by the care professionals. It goes without saying that the resulting data are vulnerable to manipulation, because of the great interests of the professionals and of the organizations involved. This problem was already described by the sociologist Blau 50 years ago. It is suggested to make a new start on a small scale with only a few indicators. It is necessary to work with unobtrusive measures-e.g. undercover observations- for observing behaviour instead of asking questions. And in the second case data should be gathered by separate organizations that are completely independent. Patients deserve that the representation of their situation should be guaranteed by scientific standards.