Science strives for objective truth. However, since it is carried out by human beings, it is not immune to error. In empirical sciences, all statements (theses) that are not axioms have the status of hypotheses, lit. ‘suppositions’. In order to be useful scientific statements, they must be falsifiable; and if they are falsifiable, they can be wrong.

Nevertheless, there is a noticeable difference between theses that have withstood falsification for a considerable time and have therefore been incorporated as theorems into a theory, on the one hand, and hypotheses that have been launched by somebody but do not seem plausible to many specialists, on the other hand. To give an example: Both the Indo-European language family and the Amerind family are concepts which can be formulated as hypotheses: there is, in both cases, an enumerated set of languages which are cognate in that they all stem, in the genealogical sense, from a common ancestor language. There is, however, a palpable difference between the two hypotheses: The Indo-European hypothesis has been launched more than 200 years ago, has had to be adjusted occasionally, but has never been falsified in the sense that there would be serious doubts about the appurtenance of its subfamilies. The Amerind hypothesis has been launched by Joseph Greenberg in 1996, has been accepted by very few others and has been seriously doubted by many in the sense that they have not seen sufficient arguments for the appurtenance of several branches of the family.

Such a difference between two scientific hypotheses is, in the first place, a practical one: The oft-confirmed and never falsified hypothesis is considered a theorem of theories of relations among languages on the globe. One is freed of the necessity to argue for such a theorem if one wants to presuppose it in some research. Contrariwise, the doubtful hypothesis is not incorporated into accepted theories. If somebody wants to build on such a hypothesis, he would have to argue extensively and even then runs the risk that his research will not be accepted.