The problem is here to examine the validity of hypotheses, which seem always valid for some researchers, while others do not agree with them. We discuss this validity and when it seems incorrect we propose some new approaches. For behaviour genetics, we will have to go back to the original work of Fisher in 1918 and its principles, which are always used by many researchers, as well in population genetics as in demography and numerous social sciences. A paper written with Atam Vetta in 2003 discusses its hypotheses and foundations. We show that the polygenic model of Fisher and Wright, which uses also the heritability concept, is based on the hypothesis that all genes segregate independently: now we know that all genes on a chromosome, except for mishaps, segregate together. In the same paper we show also a number of different problems linked to this model. We propose then new concepts: the first one could be the species value of a gene, another is regulatory genes. Molecular genetics is the key to verify the value of these hypotheses. In paleodemography the models used to estimate age at death of populations without vital records, when we have only biological indicators, remain disappointing. A paper written with Henri Caussinus for Population in 2010 (Estimating age without measuring it: a new method inpaleodemography) shows clearly these problems. The methods we propose is fully Bayesian, as commonly understood in statistics. They fit into a clear statistical environment that allows a proper validation of its theoretical properties. The simulation-based comparison with earlier methods and an application to actual archeological data show the clear superiority of our approach. More detailed analyses undertaken with the paleodemographers Isabelle Séguy and Luc Buchet permit to confirm its validity.