[Harmless: fact or fiction?].

The aim of this article is to critically discuss what conclusions can be drawn about health risks or innocuousness of exposures from an epistemiological point of view. For the interpretation of the epidemiological evidence of health risks, several assumptions have to be met--some of them cannot be empirically checked. Replication of study findings does not necessarily imply causation. However, it is possible to falsify hypotheses based on emipirical studies. This falsification principle can also be used for the interpretation of a single study. If we are able to falsify all non-causal hypotheses, the causal hypothesis is the only hypothesis that has survived and therefore should be taken seriously. Regardless of the number and kind of empirical studies, we cannot logically prove that an exposure has a harmful effect or not. Therefore, a hypothesis that could not be falsified until now should be used for political or other decisions even if there is a chance that this hypothesis will be falsified in the future when based on more appropriate trials of falsification.